Volume 5, Issue 8, August – 2020 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Real Time Attendance Marking System

Riya John Akhilesh.s Gayathri Geetha Nair
Mar Baselios college of engineering Mar Baselios college of engineering Mar Baselios college of engineering
and technology Asst professor, and technology Dept of ECE and technology Dept of ECE
Dept of ECE Trivandrum,India Trivandrum,India
Trivandrum,India

Jeen Raju Krishnendhu.B

Mar Baselios college of engineering and technology Mar Baselios college of engineering and technology
Dept of ECE Dept of ECE
Trivandrum,India Trivandrum,India.

Abstract:- Attendance management is an important on. While using RFID tag masquerading is easy for
procedure in an educational institution as well as in employees and in other method attendance of a group of
business organizations. Most of the available methods people cannot be marked at the same time.
are time consuming and manipulative. The traditional
method of attendance management is carried out in Attendance marking is effective only when it is real-
handwritten registers. Other than the manual method, time. The disadvantage of the previous methods can be
there exist biometric methods like fingerprint and trimmed down by employing facial recognition to mark the
retinal scan, RFID tags, etc. All of these methods have attendance. Face detection, which is a simple and easy task
disadvantages, therefore, in order to avoid these for humans is not so for computers. It has been regarded as
difficulties here, we introduce a new method for the most complex and challenging problem in the field of
attendance management using deep learning computer vision due to large intra-class variations caused
technology. Using deep learning we can easily train a by the changes in facial appearance, lighting, and
data-set. Real-time face algorithms are used and expression. Such variations result in the face distribution to
recognized faces of students in real-time while attending be highly nonlinear and complex in any space that is linear
lectures. This system aims to be less time- consuming in to the original image space. Face detection is the process of
comparison to the existing system of marking identifying one or more human faces in images or videos.
attendance.The program runs on anaconda flask So we aim at making the attendance- taking process
server.Here real time image is captured using mobile automatic and hands- free by employing facial recognition
phone camera. The faces on the image of the persons systems so it can automatically update digital attendance
are then recognized and attendance is marked on an records in real- time. This will computerize the traditional
excel file. method of attendance marking and thereby provides a better
user interface. This can be achieved by using a technology
Keywords:- Flask Server, Face Detection, Face called deep learning, which make use of neural networks to
Embedding, Face Recognition. teach computer by observing data. The proposed system
capture live images using mobile phone cameras and the
I. INTRODUCTION face detection is done using a frontal face detector. The
detected face are then converted to face embedding using
Nowadays, technology has improved a lot but still, facenet[6][9].A neural network is used for training process
attendance management is a major issue. Attendance where 40 to 50 images per person are face embedded and
marking is necessary to evaluate the performance of the trained. The detected face embedding are then predicted
students and the employees. There exist many methods of from the trained model for recognition process. This is then
attendance marking. The traditional method involves the marked to an excel sheet as attendance.
signature on paper sheets or manual systems that require
human sight. Many methods have been developed in order II. PREVIOUS WORKS
to improve the existing method. Considering the case of a
class, taking manual attendance every day and every period One of the methods was to mark the attendance was
is such a tedious, time-wasting process and requires a lot of by using a motion sensor and a camera which turns on when
human effort. Here the attendance will be carried out in the motion is detected by the sensor. Here the process is that
handwritten registers. It causes interruptions in the class, when the sensor detects motion it automatically makes sure
very error-prone and wastage of lot of resources. Attendance the camera is recording a video of the students and
management is not only a problem within the classroom, but compares the input data with the data that is stored in the
it also affect the business organization. Other than the database. The defect of this system is that even the slightest
manual method, there exist other methods that is biometric movement will result in the camera to turn on[1]. The
methods like fingerprint, retinal scan, and RFID tags and so second method is based on deep learning face recognition

IJISRT20AUG617 www.ijisrt.com 1112

Volume 5, Issue 8, August – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
where using Dlib which analysis the chin and nose of the IV. PROPOSED SYSTEM
person and compares it with the Rfid tag logged at that
moment.This is a tedious task and takes a long time for  Server
processing[2]. The third method is the multi-face To give the input from the phone camera a server is
recognition where faces were detected for straight faces and created in anaconda flask server. The server here is a
faces at a distance of about 200cm. This process can be computer program that provides functionality for other
implemented to get the most faces at the point of time[3]. programs or devices clients(mobile phone).The server
The fourth method is face recognition using raspberry pi program is combined with a face recognition model .While
where the face is recorded on the camera and compared the program is running on the system, by browsing the
with the stored data and the value preset and absent is server IP address we can access the phone camera as shown
marked according to the recognized face. This process can in figure 2.The IP address identifies the location of a
only recognize a few people at the time and due to lighting specific server that is connected to the internet. Here the
issues, it can’t recognize a large number of people in the mobile phone and the system running the program is
class[4]. connected to the same wifi.By choosing the camera option
live image can be captured as shown in figure 3.This image
III. REAL TIME ATTENDANCE MARKING is then uploaded as shown in figure 4 for further face
SYSTEM detection and recognition

The proposed system workflow is as shown in Figure

1. In order to capture live image using mobile phone camera
a web server is created. This image will be the input to the
face recognition model. The server program which is
combined with the face recognition model will be running
in anaconda flask server.On browsing the server Internet
Protocol (IP) address on mobile phone a window opens that
enable the access to phone camera and capture the live
image.This image is uploaded for further face recognition
process,where faces are detected and their embeddings are
created using facenet[8][9].The embedding values of the
detected face are then predicted from the embedded values
of trained data. After analyzing the data it marks the
attendance with the respect to identifying faces in the
images with values like a present and absent. This digital
information is marked automatically to the name of the
student without external help.

Fig 2:- Accessing phone camera

The live image captured for recognition is shown as in

the figure 3
Fig 1:- Proposed Work-flow

IJISRT20AUG617 www.ijisrt.com 1113

Volume 5, Issue 8, August – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
In this cascade of classifier instead of applying all
features on a single window, it groups the features into
different stages of classifiers and then applied one by one.
The first few stages of the classifier contain very less
features and if a window in this stage fails, then remaining
features are not considered. If it passes, the second stage of
features are applied and this process is continued as shown
in figure 6. For the detected frontal faces their face
embeddings are created using facenet which is then used
for face recognition[6][9].

Fig 3:- captured image

Fig 4:- uploading the image

 Face Detection
The uploaded image is given to the pretrained
classifier, that is the haarcascade frontal face detector for
the face detection process. It is a Haar feature based cascade
classifier proposed by Paul Viola and Michael Jones [5].
This classifier is trained from a lot of positive(images of
faces) and negative images(images without faces) and then
used to detect objects in other images. They uses some
windows as shown in figure 5 to identify the facial Fig 6:- Face Detection Process
features.
 Face recognition
For the training process,40 to 50 frontal face images
of each person are used. Here we used images of 7 persons
for training purpose as shown in figure 7.

These images are converted into face embedding

using facenet[6][9] and trained using neural network.

This trained model weights is then used for face

recognition. Here 67% of total image is used for training
and rest is used for testing. Detected face embeddings of
the input image is then predicted from trained model for
recognition .Faces with similar values are identified as
same persons.

Fig 5:- Haar Cascade Frontal Face Detector

IJISRT20AUG617 www.ijisrt.com 1114

Volume 5, Issue 8, August – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
VII. CONCLUSION

This method of Real-time attendance marking system

based on deep learning is to overcome the disadvantages of
the previous works which include the delay of the
processor, accuracy, etc. Deep learning extracts the features
of the input dataset and this input dataset is trained using
the neural network and this can be done using different
deep learning platforms such as Keras, tensor flow and so
on. The machine learning program can be done using a
platform called Anaconda. Anaconda is free and open
source distribution of the python and R programming
Fig 7:- Folder with trained mages languages for machine learning etc. Anaconda provides a
flask server where the server program combined with face
 Marking attendance onto the excel recognition model is deployed.Using the server IP
After the process of face detection,the embedding address,the browser connects mobile phone to the web
values are then used to predict whether the person on the server. In this attendance marking system, the live images
input image and the trained dataset are similar .In the face are taken using mobile phone camera via server and then
recognition process the known and unknown persons are uploaded for face detection and recognition. In this method
marked as present or absent in the excel sheet.In the Figure images are converted into face embedding and prediction is
8 we can see that the faces are recognized and marked done. Known and unknown faces are then marked as
properly. present and absent onto an excel sheet.

REFERENCES

[1]. M. Yadav, A. Aggarwal and N. Rakesh, "Motion

Based Attendence System in Real- Time Environment
for Multimedia Application," 2018 8th International
Conference on Cloud Computing, Data Science &
Engineering (Confluence), Noida, 2018, pp. 332-336.
[2]. X. Qu, T. Wei, C. Peng and P. Du, "A Fast Face
Recognition System Based on Deep Learning," 2018
11th International Symposium on Computational
Intelligence and Design (ISCID), Hangzhou, China,
2018, pp. 289-292
Fig 8:- Attendance Marking [3]. T. Mantoro, M. A. Ayu and Suhendi, "Multi-Faces
Recognition Process Using Haar Cascades and
V. SYSTEM SPECIFICATIONS Eigenface Methods," 2018 6th International
Conference on Multimedia Computing and Systems
This method uses neural network for training .The (ICMCS), Rabat, 2018, pp. 1-5.
computer language to be used is python. The platform used [4]. M. Arsenovic, S. Sladojevic, A. Anderla, and D.
is anaconda and flask server. Flask is a web framework that Stefanovic, "FaceTime — Deep learning-base face
provides with tools ,libraries and technologies that helps to recognition attendance system," 2017 IEEE 15th
build a web application in python. International Symposium on Intelligent Systems and
Informatics (SISY), Subotica, 2017, pp. 000053-
VI. RESULT 000058.
[5]. P. Viola and M. Jones, "Rapid object detection using a
Input has been taken from the mobile phone camera boosted cascade of simple features," Proceedings of
which helps in the real time attendance management.Faces the 2001 IEEE Computer Society Conference on
from the input image are identified using haar features and it Computer Vision and Pattern Recognition. CVPR
is converted to face embedding .In training process at first 2001, Kauai, HI, USA, 2001, pp. I-I, doi:
the imaes are converted to face embedding from which 67% 10.1109/CVPR.2001.990517.
of total image are used for training and rest is used for [6]. F. Schroff, D. Kalenichenko and J. Philbin, "FaceNet:
testing purpose.During the recognition process these A unified embedding for face recognition and
embeddings are compared and the faces are recognized the clustering," 2015 IEEE Conference on Computer
attendance is successfully marked in the excel sheet. Vision and Pattern Recognition (CVPR), Boston, MA,
2015, pp. 815-823, doi:
10.1109/CVPR.2015.7298682.

IJISRT20AUG617 www.ijisrt.com 1115

Volume 5, Issue 8, August – 2020 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[7]. G. Lou and H. Shi, "Face image recognition based on
convolutional neural network," in China
Communications, vol. 17, no. 2, pp. 117-124, Feb.
2020, doi: 10.23919/JCC.2020.02.010.
[8]. Z. Yu, F. Liu, R. Liao, Y. Wang, H. Feng and X. Zhu,
"Improvement of Face Recognition Algorithm Based
on Neural Network," 2018 10th International
Conference on Measuring Technology and
Mechatronics Automation (ICMTMA), Changsha,
2018, pp. 229-234, doi:
10.1109/ICMTMA.2018.00062.
[9]. William, D. R. Ignatius Moses Setiadi, E. H.
Rachmawanto, H. A. Santoso and C. A. Sari, "Face
Recognition using FaceNet (Survey, Performance
Test, and Comparison)," 2019 Fourth International
Conference on Informatics and Computing (ICIC),
Semarang, Indonesia, 2019, pp. 1-6, doi:
10.1109/ICIC47613.2019.8985786.
[10]. F. Schroff, D. Kalenichenko and J. Philbin, "FaceNet:
A unified embedding for face recognition and
clustering," 2015 IEEE Conference on Computer
Vision and Pattern Recognition (CVPR), Boston,
MA, 2015, pp. 815-823, doi:
10.1109/CVPR.2015.7298682.
[11]. W. Min, M. Fan, J. Li and Q. Han, "Real-time face
recognition based on pre-identification and multi-scale
classification," in IET Computer Vision, vol. 13, no.
2, pp. 165-171, 3 2019.

IJISRT20AUG617 www.ijisrt.com 1116