Automated Attendance Systems Using Face Recognition by K-Means Algorithms

3URFHHGLQJRI,QWHUQDWLRQDO&RQIHUHQFHRQ6\VWHPV&RPSXWDWLRQ$XWRPDWLRQDQG1HWZRUNLQJ
Automated Attendance Systems Using Face

Recognition by K-Means Algorithms
Dr. PALANIVEL .N
Associate Professor,
Department of Computer Science and Engineering,
Manakula Vinayagar Institute of Technology,
Puducherry.
npalani76@gmail.com
ASWINKUMAR.S BALAJI.J
Department of Computer Science and Engineering, Department of Computer Science and Engineering,
Manakula Vinayagar Institute of Technology, Puducherry. Manakula Vinayagar Institute of Technology, Puducherry.
aswinkumars2020@gmail.com balaji.devi17@gmail.com
ABSTRACT: algorithms that take a picture as the input and come back the
Attendance plays a crucial role in educational institutions, parameters associated with the image. The image has bound
several of the other industries and workplaces. Nowadays features like image improvement, segmentation, image
it's taken by standard methodology i.e. Attendance are restoration, compression, etc. Image process could be a kind
taken manually. This method takes a great deal of your of signal process that takes associate input within the kind of a
time and additionally there might be an error. Face picture from video or CCTV footage and processes the input
recognition system could be a technology capable of by using varied image process techniques and returns the
distinguishing or confirming someone from a digital image output within the kind of a picture or characteristics of an
or from a video supply. In this paper, we tend to build a image like size, skin, eyes, nose, mouth, hair. Using these
system that marks the presence of students or employees characteristics we tend to might detect the face of someone by
by recognizing their faces and manufacturing the comparison with features kept within the database.
attending sheet automatically. Face Recognition's Attendance plays a vital role in find a student's educational
accuracy rate is definitely littered with the factors like performance. It shows however the students are participating
changes in illumination, posture changes, expression in academics. This can be one of the vital challenges faced by
changes, and occlusion. In this paper, a K-means all the schools, colleges and even within the workplaces of
clustering algorithmic rule is employed to research the varied organizations. In past days attending was taken
facial expression. The biometric features of the face unit manually i.e. by calling the names of each and every student
are extracted and also the K-mean clustering technique is or by passing the attending sheet to all the students one by
used to cluster the face features. Then, SVM methodology one. But the most disadvantaged of using this standard
is employed to classify the features of the image. It may methodology is that it takes tons of your time, and there's an
accomplish high recognition performance with fewer opportunity for the incidence of miscalculation. It’s terribly
feature numbers. Finally, a report (attendance sheet) is tough to retrieve data by using the standard methodology.
generated for interpretation.
As technology grows, attendances are stored in a database
Keywords: Face recognition; Attendance sheet; Image with security. All the information concerning the students is
features; illumination; posture; expression; intensities; kept within the database. So that Manual attending can be
color; modified K-means algorithms; biometric features; replaced by the RFID [20][21].This is primarily based
SVM; Classify; interpretation attending system within which a tag (passkey) is provided that
would be utilized by the students to cast their attendances. The
I. INTRODUCTION tag scanner is used to read the tag and automatically update
Technology is developing very fast and everything is group action for every specific student. As it stores the
obtaining automatic nowadays, so as to cut back the human information within the database, retrieval of data becomes
effort. Within the development of Database Technology, terribly simple. However, it's an obstacle that is it takes tons
Image processing plays a vital role. It’s the study of of your time i.e. every student needs to wait within the queue
1
#,(((
to cast his attending. There’s an opportunity of creating of activity face detection with less number of features. The
miscalculation too that single student could carry multiple tags most disadvantage is that, below non-central and non-
and will mark attendances for multiple students. In order to coordination conditions, the face recognition continues to be
beat the drawbacks of the RFID [20][21], the biometric difficult.
authentication was introduced. Biometric authentication
includes thumb impression, iris pattern and face detection. In 2017 Li, Xiang-Yu analysis team projected a face
Thumb impression is that the major methodology that's identification methodology supported fast Principle
employed by numerous organizations for varied security and Component analysis so as to beat the matter of not obtaining
authentication functions like Aadhar card, authentication in correct lead to low face recognition beneath non- limiting
banks, phone locks and Several corporations use for condition. They used the Haar-feature classifier to extract the
maintaining attendances of the employees. Fingerprint reads feature from original database and addition to that the PCA[5]
the pattern of the finger and stores like a picture within the methodology is employed to the method for extracted
database. The stored image is processed so as to extinguish features. The LFW database is employed for verification and
features of the finger pattern from features stored in binary it provides a smart result. Similarly, within the same year,
format within the database. Since the fingerprint is exclusive Vyas and Low analysis team used the PCA [5]methodology to
for every and each person, this sort of identification is spot and analyze the facial features. The present analysis
employed. It additionally has the disadvantage of like once an techniques show that non-public identification may be
injury happens to the corresponding finger then the pattern achieved using facial expression. However a modification in
wouldn't match, and additionally obtaining the fingerprint of light, posture, scene, etc might alter the facial feature.
another person and doing identification is also possible.
In 2017, Ramakrishnan analysis team projected a paper of
Another methodology of biometric authentication is that the attendance automation using facial recognition and biometric
Iris scanner that is that the best identification methodology authentication. It principally uses Arduino and PCA[5]
employed in recent times. Humans have a novel eye pattern algorithms that may be accustomed to establish the face of the
that's used for identification or authentication purpose of a scholar who is getting into the category through a surveillance
specific person. It’s employed by the Government within the camera. It maintains the database of the scholar by noticing
method of Aadhar card implementation. However, the value the time-in and time-out by the scholar within the class and
of implementation and maintenance is incredibly high and updates the attending in keeping with it. Equally in 2013
additionally, it's just like the RFID [20][21] methodology that Arulogun O.T and analysis team projected a paper on
took tons of your time. To overcome all the on top of RFID[20][21] based mostly student attending maintaining
disadvantages, face recognition is that the best means. It is system. in this paper, they followed a technique of
used to determine whether a specific person is present in a RFID[20][21] i.e. every student was given a tag like a pass
place or not automatically with less processing time. In this that he must show to the tag reader to update the knowledge
paper we are using the K-means Clustering algorithms[1][11] like student name, class and time within the database.
[10]to seek out the presence of a specific student within the However, it's the disadvantage within which one student could
class. carry multiple tags and will update attending for over one
student.
II. LITERATURE SURVEY
In 2017, C.B.Yuvaraj and team projected an approach to
The two main elements of face recognition are feature maintain attendance using image processing techniques. The
extraction and classifiers. There are several ways employed by paper was accustomed to observe the face employing a Haar-
the researchers for getting high accuracy within the results of feature supported the Viola-Jones methodology. It finds the
face recognition system since the 1960s. Features of face face of a specific person from the video clip. It provides
identification embrace relative position of eyes, nose, mouth, results with restricted coaching time for the database set. Haar
and chin. There are several feature analysis ways. The most cascade is extremely helpful for characteristic faces, however,
vital ways among them that are in use these days are Principle changes have to be done supporting the number of cameras
Component Analysis (PCA)[5], Linear Discriminant Analysis and length of the video. Similarly, within the same year,
(LDA)[4] and Independent Component Analysis (ICA). as an Hemantkumar Rathod and team projected a technique of
example, in 2018 Pengcheng web analysis team projected a Automatic attending system uses machine learning approach.
hunt on face feature extraction supported K-means Clustering in this paper, they found the face of the scholar uses HOG
algorithms[1][11]. That was accustomed to analyzing each the features together with SVM classifier.
face and biometric features of the image. an SVM formula is
employed for classification of features. It’s achieved the target
2
#,(((
In 2007 S.Kadry and analysis team projected an approach to bunch of students sitting within the category. From that cluster
the attendance management system using iris recognition. In image, every face is known and cropped and keep as separate
that paper, iris pattern was accustomed establish the presence image . Then every cropped image is compared with the
of a specific student within the class since iris pattern is image keep within the database. If the compared image
exclusive for everybody. But the disadvantage is whereas the matches, it returns the result as successful or unsuccessful.
number of students increased day by day, the value of Then the stand out sheet is updated in step with the presence
implementation became quite troublesome. In 2001 B.Bhanu or absence of the students with the corresponding names. The
and team have given an approach for Feature extraction of the stand out document is stored within the format of that current
fingerprint image. That paper explained the feature extracted date and time for straightforward retrieval.
from the fingerprint image. The fingerprint could be a basic
methodology that's followed in several organizations to spot a BLOCK DIAGRAM OF PROPOSED SYSTEM
specific person. However, it's nearly sort of a standard
methodology wherever every student must verify himself by
putting the hand within the fingerprint detector.
In 2018 Gowtham Sandeep Nainala and team projected a

paper on SAMS Smart Attendance Monitoring System. In
that paper, they automatic the standard attending system by
uses Deep learning. The specialty of this technique is that it's
able to observe a face with completely different facial
expressions, lighting, and cause. Just in case if the system
loses to observe each face within the room there's a separate
area that takes care of it. They need to be created a
transportable device so they will take the attending even once
the category goes on without perturbing the category.
III. PROPOSED SYSTEM

Attendance plays an important role in schools, colleges, and
organizations. The standard technique of taking attending
takes longer. It’s not correct and secured. Moreover, retrieval
of knowledge uses standard technique is sort of troublesome. Fig1:Block diagram
Then there come many ways like RFID[20][21], Fingerprint
and iris scanner for attending system. However, all the The diagram consists of the subsequent stages:
strategies need voluntary action by the corresponding student
to try and do it manually. It additionally takes longer and there Stage 1: The dataset concerning the students like photos are
is also an opportunity for misusage. As technology evolves, collected from the students and preprocessed.
everything gets machine-driven. In this paper, we've planned
an inspiration for an automatic attending system by face Stage 2: From the dataset got from the scholar it locates the
recognition uses K-means Clustering.[1][11] This approach native features for every image like ears, nose, eyes, etc.
uses the face recognition by the K-means Clustering[1][11]
[10] to update the attending of a student present within the Stage 3: When locating the native features of the image, a
category automatically. This technique consists of primarily 2 facial expression of image are extracted for higher clustering.
sections - the training section and testing phase. In this stage, we are able to use the Gabor transform for
denoising.
In the training section, all the database concerning the scholar
like name, class, department, and year together with the Stage 4: When extracting the facial expression of the image,
photos are collected and store within the database. Uses this clustering of features can be done by uses modified K-means
dataset of the scholar K-means Clustering algorithms[1][11] Clustering algorithms.
are used to cluster the image. Then SVM Algorithm is trained
to predict the face uses these features. Consistent with the data Stage 5: Using the extracted and clustered features coaching
keep within the database excel sheet is generated with the can be tired order to find the face of the actual student
name of the scholar present within the category. Within the properly by SVM classifier.
testing section, the input image is taken that carries with it a
3
#,(((
Stage 6: Uses the trained dataset it creates a standout sheet
that contains the names of the students whose names are kept
within the database. The stand out sheet consists of names of
the students with attendances column left blank.
B. K-MEAN TECHNIQUE
Stage 7: Input image is taken from the schoolroom and every
The K means an algorithmic rule is comparable to distance
face within the image is detected, cropped and keep as a
primarily based algorithms. The similarity is evaluated uses
separate image.
the space between 2 points. If the space between the 2 points
Stage 8: Those separated pictures are compared with the is larger then the similarity is smaller otherwise similarity is
image keep within the database and the stand out sheet is greater. Finally get one by one class.
updated indicating the presence or absence of a specific
Divide I cluster of objects into K point teams. Calculate the
student with the corresponding name in it. Then it's stored
everyday price of data objects near to K purpose to push high
with date and time for future use. This paper primarily uses 3
similarity of purpose teams. The average value of the point
algorithms particularly SVM classifier technique, modified K-
group may be calculated as follows
means Clustering technique[1][11], and Gabor transforms
smoothing technique.
A. GABOR
The gabor/Gabor/Dennis Gabor/physicist transform
smoothing operate was planned by Gabor in 1964. it's
translation perform of mathematician function in frequency
domain. gabor/Gabor/Dennis Gabor/physicist transforms
carries with it the windowed Fourier transform Gabor function
will extract features of various direction and scales. Together Where,
with Gabor function is comparable to human eyes biological
perform. ci- nearest purpose cluster in data points I and K,i=1…n
The 2nd Gabor function is derived as Uk-center of the point of the cluster
ALGORITHM OF K-MEANS CLUSTERING

This algorithmic rule works as follows, presumptuous we've
input x1, x2, x3,….,xn and values of k
Step 1: decide k random points as cluster centers said as

In order to produce a lot of data for clustering and totally centroids.
cover frequency space, this paper among eight positions
twenty frequencies are chosen for transformation. so main
v=0~19 and direction u=0~7.
Step 2: Assign every xi to the closest cluster by calculative its
The KJ within the formula is calculated as distance to every center of mass.
Step 3: notice a replacement cluster center by taking the

common of the assigned points.
According to the formula the face feature purpose x satisfying

the distribution p(x) may be obtained .if the feature purpose
coordinate of x is (a, b), appreciate purpose x the Gabor Step 4: Repeat step two and three till none of the cluster
perform transformation formula is assignment modification.
4
#,(((
HARDWARE REQUIREMENTS:
Surveillance Camera:
x Format-1MegaPixel/1080P
x Resolution (pixels)-1280x720
x Pixels:921,600
x Aspect ratio-16:9
The computer system for executing programs:
x RAM -6GB
Fig: a x Internal memory-500GB
x Processors-Intel i3/i5
C. SVM
SVM is principally used for classification purpose. so as to SOFTWARE REQUIREMENTS:
avoid procedure complexness, it uses recognition tools by
high dimension. There are several studies using SVM as a x Python
classifier in face recognition analysis. x Deep learning
x Machine learning Algorithms
In this paper SVM classifier technique is employed, the input x HTML, JavaScript, CSS
feature is taken by a face feature vector, hyperplane is drawn x Pandas
in order to separate the two different image/features. Tthis x Numpy
hyperplane is drawn by using linear method and poly method. x File System/Mongo DB
V. VARIOUS MODULES OF THE SYSTEM

This system consists of assorted modules that may be
accustomed to implementing the application. They are
x Local Feature Extraction

x Clustering of the native feature
x Training the features
x Testing the model
x Result interpretation
A. LOCAL FEATURE EXTRACTION
The face contains several features some that modification

according to environmental changes, whereas some might not.
Features of the face embody eyes, ears, nose, mouth, etc. so as
to extract these features we want to find the native features of
Fig: b the given image within the native Feature Extraction module.
B. CLUSTERING OF A NATIVE FEATURE
IV. COMPONENTS USED IN THE SYSTEM After locating facial expression from the image individually,
These devices accommodate each hardware and code parts. all the features are clustered mistreatment the set herein
Varied operations are performed on those parts. Surveillance technique. In this module, cluster operate takes place. It may
camera and also the notebook computer are the most parts on be enforced mistreatment Gabor transform that is employed to
the hardware aspect. Moreover, Python and markup language denoise the image.
are required on the code aspect to develop the appliance.
5
#,(((
C. TRAINING THE FEATURES B. ATTENDANCE SHEET
Using the features extracted within the higher than a model. Stand out sheet is generated with the names of the students
SVM Classifiers[1][11] are trained. Since it's trained with the keep within the database when extracting and pre-processing
features of the actual image, it will be ready to predict the face the features of every and each image.
simply from the input image. Coaching may be done to the
features of the image based on condition that the data is kept
within the database.
D. TESTING THE MODEL
Once coaching is completed with the features, it

acknowledges the face from the given image. Input image
consists of a bunch of students. It locates all the faces present
within the image and stores it as a separate image within the
folder. These separate pictures are compared with the image
keep within the database. Once the compared features are
matched, it returns the result as fortunate or unsuccessful.
E. RESULT INTERPRETATION
Once the coaching and testing module is completed, it updates

the stand out sheet. Stand out sheet consists of the name of the
students within the category together with the status of their Fig. 3: Standout Sheet before Processing
attendances. When changing the attendances within the stand
out sheet, the document is saved with current date and time in C. INPUT IMAGE:
order that it may be used for the long run purpose.
Sample input image was taken within the category
VI. EXPERIMENTAL RESULT
A. STUDENT DATASET
Data Collected from the scholar is keeping within the database

is termed as the dataset.
Fig.4 : Sample Input Data taken from Class
D. CROPPED IMAGE:
Detects every face within the image and stores it individually.
Fig.2 :Student Dataset
6
#,(((
VII. CONCLUSION:
Face recognition is one amongst the identification ways. It’s a
Fig 5:Deducted Face from Input Data huge advantage among other different identification ways like
the Fingerprint, iris scanner and RFID[20][21]. All the on top
E. ATTENDANCE SHEET: of ways required a voluntary action together with bigger time
consumption. However, Face recognition is automatically
executed whereas the students are paying attention to the
category so that they could focus on the course more.
Additionally, to it, the time taken by this method is a smaller
amount. In this paper we've got planned the thought of face
recognition using the K-means Clustering algorithms, SVM
classifier together with Gabor transform and obtained a result
with 93% accuracies.
VIII. FUTURE SCOPE:

This system may be increased by adding extra database of the
students like internal marks, university communication results.
This database may be retrieved and used for a future purpose.
Further, this technique may be enforced within the field of
banking, security system, and financial aid protection. In this
paper, we've got used 3 algorithms to discover the face of the
students for group action. We tend to may additionally use
Fig6:Standout Sheet after Processing different clustering ways particularly fuzzy k-means,
mathematician Mixture Model, and hierarchical clustering.
IX. REFERENCE:
[1]. Pengcheg Wei, Zhen Zhou, Li Li and jiao, Research on

face feature extraction based on the K-means Clustering
algorithm (journal on image and video processing 2018).
[2]. WW Bledsoe, The model method in facial recognition,
vol 15 (Panoramic Research Inc, Palo Alto, 1966), p. 47
[3]. X-Y Li, Z-X Lin, in the Euro-China Conference on
Intelligent Data Analysis and Applications. Face
Recognition supported HOG and quick PCA rule
(Springer, Cham, 2017)
7
#,(((
[4]. RA Vyas, SM Shah, Comparison of PCA and LDA maximum-margin principle and the jackknife technique.
techniques for face recognition feature based extraction Pattern Recogn. 48(12), 3983–3992 (2015)
with accuracy enhancement [J]. Int. Res. J. Eng. [19]. B Martinez, MF Valstar, X Binefa, et al., Local evidence
Technol. 4(6), 3332–3336 (2017) aggregation for regression-based facial point deduction
[5]. CY Low, ABJ Teoh, CJ Ng, in IEEE Transactions on [J]. IEEE Trans. Pattern Anal. Mach. Intel. 35(5), 1149–
Circuits and Systems for Video Technology. Multi-fold 1163 (2013)
physicist, PCA and ICA filter convolution descriptor for [20]. Arulogun O. T., Olatunbosun, A., Fakolujo O. A., and
face recognition [J] (2017) Olaniyi, O.M., “RFID-Based Students Attendance
[6]. OM Parkhi, A Vedaldi, A Zisserman, in BMVC. Vol. 1. Management System”, International Journal of Scientific
No. 3. Deep Face Recognition (2015) & Engineering Research Volume4, Issue 2, February-
[7]. Y Wen et al., in European Conference on Computer 2013.
Vision. A Discriminative Feature Learning Approach for [21]. R. Karthik, V. Kaviya Priya, E. Keerthika S. Gayathri,"
Deep Face Recognition (Springer, Cham, 2016) RFID &Face Recognition System for college youngsters
[8]. NM Kopelman, J Mayzel, M Jakobsson, et al., Clumpak: Safety Enhancement", International Journal of Advanced
A program for identifying clustering modes and analysis in physical science and Communication
packaging population structure inferences across K [J]. Engineering (IJARECE) Volume five, Issue 3, March
Mol. Ecol. Resour. 15(5), 1179–1191 (2015) 2016.
[9]. L Anderson, É Aubourg, S Bailey, et al., The clustering [22]. C.B.Yuvaraj, M.Srikanth, V.Santhosh Kumar,
of galaxies in the SDSS-III baryon oscillation Y.V.Srinivasa Murthy, and Shashidhar G. Koolagudi,
spectroscopic survey: Baryon acoustic oscillations in the An Approach to Maintain Attendance using Image
data release 10 and 11 galaxy samples [J]. Mon. Not. R. Processing Techniques, International Conference on
Astron. Soc. 441(1), 24–62 (2014) Contemporary Computing ( IC3), 10-12 August 2017,
[10]. K Wagstaff, C Cardie, S Rogers, in Eighteenth Noida, India
International Conference on Machine Learning. [23]. Rekha.E, dr ramprasad P, An efficient automated
Constrained K-means cluster with database (Morgan attendance management system based on Eigen
Kaufmann Publishers Iraqi National Congress, recognition,(2017)
Williamstown, 2001), pp. 577–584 [24]. Hemantkumar Rathod, Yudhisthir Ware, Snehal Sane,
[11]. SH Kang, B Sandberg, AM Yip, A regularized k-means Suresh Raulo, Vishal Pakhare and Imdad A. Rizvi
and multiphase scale segmentation [J]. Inverse Prob. Automated Attendance System using Machine Learning
Imaging 5(2), 407–429 (2017) Approach, 2017 International Conference on Nascent
[12]. V Vapnik, R Izmailov, Knowledge transfer in SVM and Technologies in the Engineering Field (ICNTE-2017)
neural networks. Ann. Math. Artif. Intell. 81(1–2), 3–19 [25]. Shubhobrata Bhattacharya, Gowtham Sandeep Nainala,
(2017) Prosenjit Das and Aurobinda Routray, Smart Attendance
[13]. D Singh et al., in Communication, Control and Monitoring System ( SAMS&#41: A Face
Intelligent Systems (CCIS), 2015. An Application of Recognition based Attendance System for Classroom
SVM in Character Recognition with Chain Code (IEEE, Environment, 2018 IEEE 18th International Conference
Mathura, 2015) on Advanced Learning Technologies
[14]. X Chang et al., in International Conference on Machine
Learning. Complex Event Deduction Using Semantic
Saliency and Nearly-Isotonic SVM (2015)
[15]. C Ding et al., Multi-directional multi-level dual-cross
patterns for robust face recognition. IEEE Trans. Pattern
Anal. Mach. Intell. 38(3), 518–531 (2016)
[16]. Y Gao, J Ma, AL Yuille, Semi-supervised sparse
representation based classification for face recognition
with insufficient labeled samples. IEEE Trans. Image
Process. 26(5), 2545–2560 (2017)
[17]. MS Hossain, G Muhammad, Cloud-assisted speech and
face recognition framework for health monitoring. Mob.
Netw. Appl. 20(3), 391–399 (2015)
[18]. C-C Chang, S-H Chou, Tuning of the hyperparameters
for L2-loss SVMs with the RBF kernel by the
8
#,(((

Automated Attendance Systems Using Face Recognition by K-Means Algorithms

Uploaded by

Automated Attendance Systems Using Face Recognition by K-Means Algorithms

Uploaded by

3URFHHGLQJRI,QWHUQDWLRQDO&RQIHUHQFHRQ6\VWHPV&RPSXWDWLRQ$XWRPDWLRQDQG1HWZRUNLQJ