0% found this document useful (0 votes)
14 views8 pages

Face Recongnition Attendance System Using Machine Learning

The document states that the training data extends up to October 2023. It implies that any information or developments occurring after this date are not included. This sets a clear boundary for the relevance of the data.

Uploaded by

jethalalgada9325
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
14 views8 pages

Face Recongnition Attendance System Using Machine Learning

The document states that the training data extends up to October 2023. It implies that any information or developments occurring after this date are not included. This sets a clear boundary for the relevance of the data.

Uploaded by

jethalalgada9325
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOC, PDF, TXT or read online on Scribd
You are on page 1/ 8

FACE RECOGNITION ATTENDANCE

SYSTEM USING MACHINE LEARNING


Tejas Janjire1, Om Dagade2, Sai Jagdale3, Sanika Jawanjal4, Prof. A.M. Jagtap5,
1,2,3,4,5
Department of Computer Engineering, All India Shree Shivaji Memorial Society College of Engineering, Pune
Abstract - Attendance monitoring stands as a pivotal aspect across educational institutions, corporate environments,
and various organizational settings. Traditionally, attendance is recorded manually, often through verbal calls or
signature-based methods, which prove to be time-consuming and prone to errors in maintaining accurate records. With
regard to this, the Recognizing the inefficiencies inherent in manual systems, there arises a demand for automated
solutions. In this proposed study, the focus is on the integration of facial recognition technology to streamline
attendance tracking processes. The suggested approach entails the utilization of sophisticated technique for face
detection and recognition. Specifically, the methodology incorporates the Histogram of Oriented Gradients (HOG)
algorithm for precise face detection, coupled with Deep Convolutional Neural Networks (CNN) for robust face
recognition capabilities. By harnessing these advanced techniques, the aim is to develop an automated attendance
system that enhances efficiency and accuracy in record-keeping. Further the attendance is pushed on the server, the
students and the teachers can access it from the website.
Keywords - Attendance System, HOG, CNN, Facial Landmarks, Open CV, Face Recognition.

I. INTRODUCTION

Problems associated with manual attendance can be resolved by implementing an automated biometric face recognition
system. In recent years, various biometric systems for attendance management have been proposed, including fingerprint
scanning, iris scanning, and hand scanning. However, these systems have drawbacks in terms of accessibility, portability,
and, most importantly, implementation cost.
The aim of this study is to provide an alternative method of attendance system using a facial recognition system. The
attendance achieved from this system is stored in a database accessible to teachers and administrators. In contemporary
organizational and educational contexts, maintaining precise attendance records is essential for accountability and
performance assessment. However, the task becomes arduous, especially in large classrooms or workplaces, where manual
attendance management is error-prone and labor-intensive. To tackle these challenges effectively, the integration of facial
recognition technology emerges as a groundbreaking solution. By harnessing sophisticated techniques such as Machine
Learning and Deep Learning, facial recognition systems can swiftly and accurately identify individuals in real-time. This
not only simplifies attendance tracking processes but also elevates accountability and quality assurance, particularly in
academic settings where attendance closely aligns with academic achievement.
The proposed methodology revolves around leveraging facial recognition to mark attendance with precision. Through
techniques like HOG for face detection and ResNet34 for facial recognition, the system efficiently identifies individuals as
they enter the premises. The proposed approach focuses on recognition of students faces by comparing the student faces
from the database and marking the attendance of the student whose face is recognized in the excel sheet. After generating
excel sheet, system upload the excel sheet on the website. Then every student will be able to see the attendance records in
their login and admin will be able to see all student records.
Moreover, the implementation includes automated email notifications to alert parents of defaulter students, promoting
parental engagement and transparency. This innovative approach not only overcomes the shortcomings of traditional
attendance monitoring but also underscores the significance of embracing technological advancements to streamline
organizational workflows in contemporary settings.

II. LITERATURE REVIEW


1. Attendance Management System using Facial Recognition [1]: This system streamlines attendance processes through a
combination of HOG and neural network algorithms. Its workflow comprises face detection, pose estimation, facial
encoding, and identification.

2. Smart Attendance System using Dlib Pre-Trained Neural Network [2]: This solution addresses face recognition
challenges using HOG and SVM classification techniques. It accurately identifies users and logs attendance directly
into a Google Sheet, boasting a remarkable 95.23% accuracy rate.

3. Real Time Face Recognition Based Attendance System for University Classroom [3]: This system records attendance
and also archives data securely in an Excel sheet. Initial image capture in HD 1080p, followed by noise reduction and
HOG-based facial feature detection, leads to a face recognition accuracy of 77.38%.

4. Non-Contact Temperature Detection, Face Mask Detection, and Attendance Updation System using Facial Recognition
Technique [4]: Departing from traditional card swiping or biometrics, this system enhances safety by dispensing
sanitizer upon entry. Using the HOG algorithm, it achieves accuracy rate of 96.67% while detecting faces and updating
student attendance.

5. Design of E-Attendance Checker through Facial Recognition using Histogram of Oriented Gradients with Support
Vector Machine [5]: This system implements facial recognition using HOG and SVM algorithms, achieving an
impressive accuracy of 95.65%.

6. An approach towards development of automated attendance system using face detection and recognition [6]: By
combining face detection, image processing, and recognition methodologies, this system offers a comprehensive
attendance solution. Leveraging the HOG approach for detection and LBPH method for recognition, it attains an
accuracy of 91.75%.

7. Class attendance management system using face recognition [7]: Operating on Raspberry Pi with Raspbian OS, this
system captures images and employs the LBPs algorithm for face recognition. With an accuracy of 95% and a dataset of
11 persons, it provides reliable attendance management.

8. A smart technique for attendance system to recognize faces through parallelism [8]: This system utilizes classroom
images and database records for face detection and recognition. Through histogram classification, noise removal, and
other steps, it achieves automatic attendance marking based on face recognition.

9. Automated Attendance System using Machine Learning Approach [9]: Viola-Jones and HOG combined with SVM
address various real-world challenges such as scaling and occlusions. This system ensures accuracy while eliminating
redundancy in manual attendance records.

10. Automatic Attendance Management System Using Face Detection [10]: Fixed cameras capture images for face
detection and recognition against a database, allowing for automatic attendance marking. Additionally, it notifies
parents of absent students, utilizing eigenfaces for facial recognition.

III. METHODOLOGY

This section deals with the steps required for face detection and face recognition. The work includes nine phases: System
Architecture, creating a database and its training, Image Acquisition, Implementation Of HOG, Detection of Facial
landmarks, creating face encoding, finding a corresponding name of a person name from the encodings, Sending attendance
records to the parent, Automatic generation of Attendance record of student on the website.
Fig 1: Flow Diagram
A. System’s Architecture
This system captures the image of students through a camera. The captured image is then fed as an input to our Face
Recognition System. After capturing the image, it is converted from BGR to RGB format. Next, the image is encoded
using Histogram of Oriented Gradients (HOG). Once the features are extracted from facial images, our face
recognition system compares them with the system’s database. It marks the attendance of all those students whose
facial landmarks match with our database. This system's major objective is to minimize processing time as much as
feasible.
B. Database Creation and Training:
Initially, the system establishes a database containing facial images of students. These images serve as the basis for
feature extraction, accomplished through a deep convolutional neural network model.
C. Image Acquisition
A high-definition camera is used for image capturing purposes in the classroom. An HD 1080p camera provides
images with high pixel quality, which increases the overall system’s accuracy.
D. Implementation of HOG
Image Pre-processing: The images undergo pre-processing, adjusting the aspect ratio to 1:2 and resizing to 64x128
pixels. Subsequently, they are divided into cells of either 4x4 or 8x8.
Gradient Calculation: Gradient values are computed in both the x and y directions, determining the magnitude and
orientation of gradients for each pixel.
Histogram Construction: A histogram is constructed based on the gradient magnitude and orientation values,
resulting in a 1x9 matrix.
Normalization: Grouping four cells together to form a 1x36 matrix, normalization is then applied to standardize
the matrix values.

E. Detection of Facial Landmarks


Sometimes, systems fail to accept facial images of students due to variations in angles, resulting in the inability to
identify students through their facial features. To mitigate this issue, it is crucial to accurately detect facial landmarks.
While the HOG method excels in frontal face detection, it encounters challenges with faces in varying orientations or
when faces are in different directions. To overcome this limitation, facial landmarks are employed. This method
involves identifying consistent points on the face such as the eyes and lips, regardless of facial direction. Utilizing a
68-point facial landmark model, which accurately detects 68 key facial points, facilitates robust facial
landmark estimation.

Fig 2: Facial Landmarks

F. Creating Face Encodings:


The process of generating face encodings is pivotal in facial recognition, enabling the conversion of facial features
into a numerical format for efficient comparison and analysis. Face encoding is a method of representing the face with
128 computer-generated measures. Two photographs of the same people would have similar encoding, whereas two
photographs of two distinct persons would have completely different encoding.
In this phase, the system employs a neural network architecture, Resnet34, for generating face encodings. Res Net,
short for Residual Network, is renowned for its exceptional performance in image recognition tasks, achieved through
the utilization of residual connections.
The face encoding process entails feeding facial images through the ResNet34 neural network. Ultimately, the
network produces a 128-dimensional vector, termed the face encoding, encapsulating the unique traits of the input
face.
Face encoding is a method of representing the face with 128 computer-generated measures. Two photographs of the
same people would have similar encoding, whereas two photographs of two distinct persons would have completely
different encoding. In this study, a face recognition model is proposed to make the basic face recognition process
simpler. Here the time-consumed for the process is reduced and the complexity of recognition is lessened. Face
encoding, feature extraction, and classification are the three basic processes in the face recognition approach.
G. Finding the Corresponding Names of person from the Encodings:
Once the face encodings are extracted, the subsequent step involves comparing them with those stored in the database.
This comparison is executed using a similarity metric, such as the Euclidean distance, quantifying the likeness
between two face encodings.
For each unknown face encoding, the system traverses the database of known face encodings, computing the
Euclidean distance between the unknown encoding and each known one. The unknown face is then paired with the
nearest known face based on the minimal Euclidean distance.
If the Euclidean distance between the unknown encoding and a known one falls below a predefined threshold, the
system identifies the corresponding individual and logs their presence in the attendance record.
H. Uploading Attendance on web server

Following the attendance marking, an Excel sheet is generated to document the attendance status of each student
comprehensively. The attendance of each student is automatically uploaded on the web server. The students and the
teacher can access the attendance record from the server on the website. The teacher can send email automatically to
the parents of the defaulter students in one click.

Comparison of various face recognition algorithms.


No Face Accuracy of Drawbacks
recognition system
method
1 Principle 92% Work for simple
component background &
analysis sufficient light
source
2 Eigen faces >85% Limitations over
the variations in
light, size & in
the head
orientation
3 Independent 81.35% ---
component
analysis
4 PCA & --- Recognition
Eigenfaces only frontal face
5 MTCNN 98.87% ---
6 Support vector 85-92.1% High
machine algorithmic
complexity and
extensive
memory
requirement
7 Dlib 99.38% ---

IV. RESULT

The result of the provided work is shown below. The final result is obtained on the website from the server. Our system
takes the input image from the camera and goes through all the steps mentioned above and finally the attendance of the
students whose face is recognized is uploaded on the server. Our system has been tested on 30 students whose faces were
first stored in the database. After the face images of the students have been stored in the database then our system is trained
for those images. While training our system stores 128 face encodings of each student in a JSON file. During the
recognition phase the student face in front of the camera is first encoded, then the encodings are matched with the images
stored in the database. The encodings are compared using Euclidean Distance based on some threshold limit (we have used
0.5 as threshold limit). The student whose face is recognized is marked as present and his attendance is uploaded directly
on the server. The teacher can access attendance of all the students from the website using the given login credentials. The
students can also see their attendance from the website using the login credentials given to them. The teacher can send
email to the parents of defaulter students in one click from the website.
Fig 3: Face Recognition

The above image shows the face recognition of multiple students in one frame.

Fig 4: Admin Console

Fig 5: Student Console


V. CONCLUSION

Hence, we have developed a face recognition attendance using machine learning. For developing the attendance system we
have used HOG and ResNet34 algorithms. The system has been tested on 50 students and the faces of all the 50 students
has been correctly recognized.

REFERENCES

[1] Sarath Krishnan P, et al. “Attendance Management System Using Facial Recognition” IEEE 2022.
[2] Nivedhitha DP, et al. “Smart Attendance System using Dlib Pre-Trained Neural Network ” IEEE 2023.
[3] Hajrah Sultan, et al. “Real Time Face Recognition Based Attendance System For University Classroom” IEEE 2022.
[4] Prajwal C Hegade, et al. “Non-Contact Temperature Detection, Face Mask Detection, and Attendance Updation System using Facial Recognition
TechniqueClass Attendance System Using Viola-Jones Algorithm and Principal Component Analysis” IEEE 2021.
[5] Allan Jason C, et al. “Design of an E-Attendance Checker through Facial Recognition using Histogram of Oriented Gradients with Support Vector
Machine” IEEE 2020
[6] Sayan Seal, et al. “An approach towards development of automated attendance system using face detection and recognition” IEEE 2020.
[7] Salim, Omar Abdul Rhman, Rashidah Funke Olanrewaju, and Wasiu Adebayo Balogun. "Class attendance management system using face recognition."
IEEE 2018.
[8] B Prabhavathi et al. “A smart technique for attendance system” to recognize faces through parallelism IOP Conference Series: Materials Science and
Engineering, 2017.
[9] Rathod, Hemantkumar, et al. "Automated attendance system using machine learning approach." International Conference on Nascent Technologies in
Engineering IEEE, 2017.
[10] E.Varadharajan et al. “Automatic Attendance Management System Using Face Detection” Online International Conference on Green Engineering and
Technologies (IC-GET) 2016.
[11] Phichaya Jaturawat et al, “An evaluation of face recognition algorithms and accuracy based on video in unconstrained factors” IEEE 2016.
[12] Raj Malik et al. “Prototype model for an intelligent attendance system based on facial identification IEEE 2016.
[13] Nilesh D. Veer et al, “An automated attendance system using video surveillance camera” IEEE 2016.
[14] Samuel Lukas et al, “Student attendance system in classroom using face recognition technique” IEEE 2016.
[15] Srivignessh Pss et al, “RFID and pose invariant face verification based automated classroom attendance system “ IEEE 2016
[16] Jennifer C. Dela Cruz et al, “Biometrics based attendance checking using Principal Component Analysis ” IEEE 2015.
[17] Priyanka Wagh et al, “Attendance system based on face recognition using eigen face and PCA algorithms” IEEE 2015.
[18] Adrian Rhesa Septian Siswanto et al, “Implementation of face recognition algorithm for biometrics based time attendance system” IEEE 2014.
[19] Mashhood Sajid et al, “A conceptual model for automated attendance marking system using facial recognition” IEEE 2014.
[20] Qasim Al-Shebani et al, “Embedded door access control systems based on face recognition: A survey” IEEE 2013.

You might also like