AI Based Student Attendance Monitoring System

Sehrish Parwez1,, Chattaniya Dev Holkar 2, , Ojaswin Narang3, Harshal adya4,, Ajay Vikram Singh 5
Amity Institute of Information Technology, Noida
Amity University, Noida, U.P., India

Abstract- popularity, they are susceptible to various

inefficiencies like human errors,
Attendance tracking is an important feature of
mismanagement of data, and vulnerability to
educational institutions to keep students
fraudulent activities like proxy attendance. To
engaged and uphold discipline. Conventional
overcome these challenges, institutions have
attendance systems, such as manual roll-call
implemented biometric systems, like
and biometric, are usually plagued by
fingerprint and RFID-based attendance
inefficiencies like time wastage, human
monitoring, which have enhanced accuracy
mistakes, and security risks. With the
but are plagued by some disadvantages,
evolution of artificial intelligence (AI), deep
including high costs of implementation,
learning, and facial recognition technology,
hygiene issues, and hardware failure. With the
automated attendance systems have become a
pace of technological changes, AI-based
more dependable and efficient option. These
attendance monitoring systems have become a
AI-driven systems employ machine learning
revolutionary innovation. The systems employ
algorithms, mainly convolutional neural
deep learning, computer vision, and cloud
networks (CNNs), to identify and recognize
computing to enable machine learning-based
student faces in real time without the need for
student identification and marking of
human intervention. Through the use of AI,
attendance. Through cameras deployed in
institutions can gain greater accuracy,
classrooms, AI-based systems take students'
minimize cases of proxy attendance, and
photos, run them through facial recognition
improve security. This research paper
software, and compare them against a pre-
discusses the shift from conventional
loaded database of student profiles. This
attendance systems to AI-driven solutions,
minimizes the requirement for manual
examining their methodologies, challenges,
intervention, cutting down substantially on
outcomes, and future prospects.
roll-call time and removing human errors.
Keywords: AI-based attendance, facial Moreover, attendance systems based on AI
recognition, machine learning, deep learning, can be connected to mobile apps and real-time
student monitoring, automation, biometric dashboards, enabling teachers and
attendance, education technology, real-time administrators to keep track of attendance
tracking, cloud computing. patterns easily. This paper discusses the
shortcomings of conventional attendance systems
I. Introduction and provides a detailed examination of AI-based
solutions, including their approaches, advantages,
Attendance management plays a significant disadvantages, and future prospects. By
role in educational institutions, ensuring that integrating artificial intelligence, machine
students maintain consistent participation in learning, and computer vision, AI-based
academic activities. Traditionally, attendance attendance systems are revolutionizing the
tracking has been done through traditional way institutions manage student attendance,
methods, i.e., verbal calling of the students' making them more reliable, scalable, and
names and recording their attendance in effective.
registers. Although such methods have gained

1
 Although this is easy and involves no extra
technology, it has a number of disadvantages.
It is time-consuming, particularly for large
classes, where attendance roll call can account
for a big chunk of lecture time. Moreover, it is
susceptible to human errors, for example,
mispronunciations, unintentional omissions, or
double entry. The problem of proxy
attendance is also inevitable, as students can
respond in place of absent classmates. Long-
term maintenance and storage of manual
Figure 1 attendance records make it cumbersome and
inefficient in handling administrative tasks.
II. Literature Review
2) Challenges with Manual Attendance
Different attendance management methods The main disadvantages of manual attendance
have been researched and implemented over
systems are:
the years, from roll call to automated
biometric systems. Studies indicate that even • Time Inefficiency: Taking attendance
though the traditional methods have worked, the old-fashioned way in big classes
they are accompanied by drawbacks that do wastes precious class time.
not facilitate efficiency and accuracy. • Error-Prone Nature: Marking
Biometric solutions like fingerprint scanning attendance by human beings can be
and RFID-based systems were developed to prone to errors.
counter these issues; however, they are still
• Proxy Attendance Risk: Students
plagued by system failure issues,
inconvenience to the users, and security may incorrectly check themselves or
issues. others present.
In the past few years, AI-based attendance • Issues in Data Management: It is
systems have emerged as popular choices with challenging to store, access, and
advancements in computer vision and deep analyze paper-based records..
learning. Research has revealed that facial
recognition technology integrated with deep 3) Drawbacks of Biometric Attendance
learning algorithms such as Convolutional
Neural Networks (CNNs) has the ability to Systems
provide over 98% accuracy rates, hence a In order to counter the limitations of manual
worthy alternative to the conventional attendance, systems based on biometrics like
biometric systems. fingerprint scanners, iris scanners, and RFID
card systems were developed. These systems
1) Manual Attendance System ensure a greater level of accuracy and
Manual tracking of attendance is one of the
minimize the risk of proxy attendance. But
oldest and most common practices in schools.
then, they also have their own set of
In this method, the teacher rolls out students'
limitations.
names, and their attendance is marked in a
• High Implementation Costs: Installation of
physical register or electronic spreadsheet.
2
biometric devices throughout an institution pictures or video streams of attending students
involves huge investment in hardware and are taken during the lecture. These
maintenance. photographs are either processed in real time
or temporarily stored to be processed
• Sanitation Issues: Touch-based biometric
afterwards. The system is capable of
systems like fingerprint readers are sanitation accommodating a number of environments,
concerns, particularly in the post-pandemic such as varying lighting conditions, facial
era. variations, and small head sways. In contrast
to biometric systems that ask students to scan
III. Methodology their fingerprints or RFID cards manually, AI-
based systems operate quietly in the
AI-Attendance Monitoring is made to be self- background without interrupting the class.
service automation for student attendance through
cutting-edge artificial intelligence, deep learning, 1.2 Preprocessing and Feature Extraction
and computer vision algorithms. Notably, not
Once the images are collected, they are
based on any manual intervention or hardware-
dependent biometrics, attendance management preprocessed to enhance their quality and
with artificial intelligence employs the usage of remove unwanted noise that may have an
face recognition technology that is used for impact on recognition accuracy. Preprocessing
identification of the students in real-time. This is done through image normalization,
system works by taking pictures or video streams adjustment of contrast, noise removal, and
from cameras installed in the classroom, conversion into grayscale for successful
preprocessing the images for better clarity, and analysis. One of the most crucial processes
employing deep learning algorithms to identify during this stage is face detection to ensure
and authenticate student identities. that only the right segments of the image—
primarily the students' faces—are being
analyzed. The system employs deep learning-
based detection methods such as Histogram of
Oriented Gradients (HOG), Multi-task
Cascaded Convolutional Networks (MTCNN),
or Deep Learning-based Single Shot MultiBox
Detector (SSD) to detect facial landmark
features, e.g., eyes, nose, and mouth. These
core features are detected to get an individual
facial representation per student and are
compared with a pre-registered database.
To further improve accuracy, advanced
feature extraction techniques such as PCA and
LBPH are utilized. These procedures assist in
reducing dimensions while maintaining
essential facial features. Advanced feature
Figure 2 extraction techniques include Local Binary
Patterns Histogram (LBPH), Eigenfaces,
Fisherfaces, and Scale-Invariant Feature
1.1 Data Collection Transform (SIFT), which help the AI model
The initial step in an AI-powered attendance recognize faces accurately under different
tracking system is the capture of student facial lighting conditions, angles, and even with
information. Cameras are placed in a strategic partial occlusion (e.g., wearing masks or
manner in classrooms so that real-time
3
glasses). To compare two students' faces, we use
Cosine Similarity:
1.3 AI Model Training and Attendance Similarity=A⋅B/∥A∥×∥B∥
Detection Explanation:
At the heart of the AI-driven attendance  A and B are face feature values.
system is the deep learning model that drives  If Similarity = 1, the faces match
facial recognition. It is trained on (same person).
Convolutional Neural Networks (CNNs), an  If Similarity = 0, the faces do not
advanced machine-learning technique match (different people).
specifically tailored to image analysis. The
training dataset comprises several images of 2. Average Recognition Time for an Entire
every student captured from varied angles and Class
lighting conditions to enhance recognition
resilience. The AI model can learn unique If the system recognizes multiple students,
facial features, allowing it to separate the we calculate total time:
different students even from cluttered Total Recognition Time=Timeavg×Total Stud
classroom environments. ents
When the AI model is analyzing a classroom ✅ Example:
photo, it first identifies and separates student  Time per student = 0.08 seconds
faces and then matches them with the pre-  Class size = 60 students
registered database of students. A similarity
score for the face detected and the known Total Recognition Time=0.08×60=4.8 seconds
images of the students is calculated based on
Euclidean Distance Matching or Cosine This means the AI system can take
Similarity algorithms. In case the similarity attendance for a full class in just 4.8
score is greater than the pre-defined threshold, seconds!
the student is identified as present.
3. Attendance Percentage Calculation

To calculate a student’s attendance

percentage:
A=(P/T)×100
Where:
 P = Number of present days.
 T = Total working days.
 A = Attendance percentage.
For example, if a student attended 85 out of
100 classes:

A=(85/100)×100=85%A

Figure 2 4. Storage Requirement for Video-Based

Attendance System

1.4 Mathematical Concepts and Formulas If the AI system records attendance videos, we
calculate the total storage needed:
1. Face Matching Using Cosine Similarity Total Storage=Bitrate×Duration

4
✅ Example: administrators to track student presence as
 Video bitrate = 5 Mbps (Megabits per they enter or leave the classroom. Unlike
second) manual methods, this feature eliminates proxy
 Recording duration = 60 minutes attendance and significantly reduces the time
Total Storage=5×60=300 Megabits spent on roll calls.
This helps in choosing the right storage The system learns to recognize students’ facial
plan. features in different lighting conditions,
angles, and even while wearing masks,
5. Security Strength of AI System making it more reliable than fingerprint or
(Encryption Strength) card-based systems. Additionally, attendance
records are stored in a cloud-based database,
If the attendance system encrypts student data, allowing teachers, students, and parents to
we calculate its security level: access reports anytime, ensuring better record-
Encryption Strength=2^N keeping and data security. Notifications and
✅ Example: alerts are automatically sent to parents and
If the system uses 128-bit encryption: teachers when a student is absent, providing
real-time updates and promoting transparency
Encryption Strength=2^128 between schools and families.
This means a hacker would need 2¹²⁸ Another advantage is multi-device
attempts to break the security, making it very compatibility, as the system can be accessed
secure. from mobile phones, tablets, laptops, and even
existing CCTV cameras, making it convenient
for educational institutions to integrate
without requiring expensive hardware.
Furthermore, the AI-powered attendance
system can be linked with school
management systems (ERP), enabling
seamless synchronization with student data,
academic records, and performance tracking.
Overall, an AI-based attendance monitoring
system is an innovative and efficient approach
to managing student presence with improved
Figure 3 accuracy, security, and convenience. It
enhances administrative efficiency, reduces
paperwork, and provides insightful analytics
IV. Key Features for educational institutions to monitor
attendance trends, address absenteeism, and
An AI-based student attendance monitoring improve student engagement.
system is a smart and automated solution that
eliminates the need for traditional roll calls
and biometric systems, ensuring accurate and V. Result Analysis and Future Scope
efficient attendance tracking. One of its most Result Analysis
important features is automated face The use of AI-based attendance monitoring
recognition-based attendance, where systems has greatly enhanced efficiency,
students' faces are scanned and matched accuracy, and security in tracking student
against a stored database, marking their attendance. Studies and practical
attendance instantly. This system operates in implementations have demonstrated that AI-
real-time, allowing teachers and
5
based facial recognition systems deliver more AI-powered attendance systems can go
than 98% accuracy, surpassing conventional beyond mere tracking to offer information on
manual and biometric attendance systems. The student participation and engagement. AI,
biggest benefit is the absence of proxy through analysis of attendance trends, can find
attendance, which has been a chronic problem patterns of students who are likely to fall
in schools. Because facial recognition behind in their studies due to consistent
authenticates individual identities in a singular absenteeism. Future systems can be integrated
manner, students cannot falsify attendance with learning management systems (LMS) and
records like they might with manual roll calls send real-time notices to teachers, parents, or
or even some biometric systems. academic counselors if a student begins
In conventional practices, teachers tend to take exhibiting patterns of deteriorating attendance.
attendance 5-15 minutes every session, and
this translates into a large amount of lost Cloud and Edge Computing for Real-Time
teaching time when accounting for several Processing
sessions in a day over a year. AI-enabled One such challenge for AI-based attendance
attendance systems, however, take mere systems is processing lag with large amounts
seconds to take attendance for an entire class. of real-time image data. The future upgrade
This saves time, and as a result, there is better can take advantage of edge computing, where
classroom management because the teacher face recognition is computed on local devices
gets to concentrate on teaching more than on (e.g., classroom cameras) rather than uploaded
administrative duties. to a cloud server. This will minimize response
time, enhance security, and eliminate
dependency on internet connectivity,
rendering AI attendance solutions more
feasible in remote and underdeveloped
regions. Mobile apps with integrated AI can
use geolocation, facial recognition, and time
checks to confirm the presence of a student,
keeping attendance records correct for
distance education programs. Universities that
provide flexible or hybrid course models will
particularly benefit from this.

VI. Conclusion
The history of attendance monitoring has
Figure 4 witnessed dramatic leaps from the
conventional manual roll calls to biometric
Future Scope and AI-based systems. Although manual
The potential for AI-powered attendance methods have been in use for ages, they suffer
monitoring systems is limitless in the future, from inefficiencies like human mistakes, time
propelled by the growth of artificial consumption, and vulnerability to fake
intelligence, deep learning, cloud computing, practices such as proxy attendance. The
and the Internet of Things (IoT). The next advent of biometric attendance systems, like
generation of attendance tracking solutions fingerprint and RFID-based recognition, was a
can be defined through a few major areas of step towards automation, but these solutions
improvement and development. were still not without issues, like hygiene
issues, hardware malfunction, and
6
vulnerability to spoofing. With the growing and deep learning algorithms are overcoming
demand for more efficient, secure, and these challenges with enhanced model
automated solutions, AI-based attendance robustness and accuracy. The use of multi-
monitoring systems have become a game- factor authentication, like facial recognition
changing alternative, revolutionizing the way coupled with voice recognition or geolocation
educational institutions and organizations tracking, is further enhancing the
monitor attendance. dependability of these systems.
AI-based attendance systems utilize deep
learning, computer vision, and cloud References
computing to process facial recognition and
automatically identify students. These systems [1] Ahmad, T., & Mohammed, F. (2021). AI-
avoid the necessity for manual intervention, Based Smart Attendance Monitoring System
providing accurate and up-to-the-minute Using Facial Recognition. International
attendance recording. In contrast to Journal of Advanced Computer Science and
conventional biometric systems, AI-powered Applications, 12(4), 57-64.
attendance solutions are non-touch, making
them a healthier and more convenient choice [2] Chen, Y., Li, X., & Wang, Z. (2020). A
in the face of global health crises. The Deep Learning-Based Automated Attendance
incorporation of AI also increases security by System Using Facial Recognition Technology.
reducing identity forgery and unauthorized IEEE Transactions on Emerging Topics in
access, making attendance data genuine, Computing, 8(3), 432-445.
tamper-proof, and trustworthy.
The use of AI-driven attendance tracking has [3] Sharma, P., & Gupta, R. (2019).
proven various important advantages. First, it Comparative Analysis of Biometric and AI-
saves a great deal of time on attendance Based Attendance Systems in Educational
marking, freeing instructors to engage in Institutions. Journal of Artificial Intelligence
teaching activities instead of administrative Research, 15(2), 89-102
ones. Research indicates that AI-based
attendance tracking cuts down on time usage [4] Singh, K., & Verma, A. (2022). Enhancing
by more than 75%, making effective use of Security and Efficiency in Attendance
valuable classroom time. Second, it improves Tracking Through AI and Blockchain
data accuracy and availability through secure Integration. International Journal of
storage of attendance records in cloud Information Security and Privacy, 10(1), 27-
databases. This enables institutions to produce 45.
timely reports, monitor attendance patterns,
and inform student performance and academic [5] Patel, S., & Rao, K. (2018). Challenges
progress with data-driven decision making. and Future Scope of AI-Based Attendance
Automating attendance tracking eliminates Monitoring Systems. Proceedings of the
human mistakes and avoids the complications International Conference on Computer Vision
of missing or misplaced records. and Machine Learning, 21(5), 112-125.
While AI-based attendance monitoring
systems have their benefits, they also have [6] Zhang, L., & Wu, H. (2023). Facial
their challenges. These include changing Recognition for Attendance Systems:
lighting conditions, image quality, and facial Advances, Challenges, and Ethical
occlusions (masks, hats, or head coverings), Considerations. Journal of Machine Learning
among others, impacting the recognition and Applications, 9(2), 211-229.
accuracy. But ongoing developments in AI
[7] ISO/IEC 2382-37:2017. Biometrics and
7
Face Recognition Standards for Automated
Identification Systems. International
Organization for Standardization (ISO)

[8] Rahman, M., & Hasan, S. (2022). Deep

Learning Models for Attendance Monitoring
Using AI and IoT: A Review. Future
Computing and Automation Journal, 6(4), 56-
74.

[9] OpenAI. (2024). Advancements in AI-

Based Facial Recognition and Ethical
Implications in Educational Technology.
Research Paper Archive.

[10] Kumar, A., & Sharma, D. (2021). A

Survey on AI-Based Attendance Systems:
Challenges and Future Trends. International
Journal of Artificial Intelligence and
Applications, 13(2), 55-78

[11] Lee, C., & Park, J. (2020). Real-Time

Face Recognition Attendance System Using
Deep Learning and IoT Integration. IEEE
Transactions on Smart Computing, 7(5), 321-
337

8
9