IRJET- Face Recognition using Machine Learning

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 04 | Apr 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 3772
FACE RECOGNITION USING MACHINE LEARNING
Ishan Ratn Pandey, Mayank Raj, Kundan Kumar Sah , Tojo Mathew, M. S. Padmini
123B.E., Department of Computer Science and Engineering, The National Institute of Engineering, Mysuru, India
45Assistant Professor, Computer Science & Engineering, The National Institute of Engineering, Mysuru, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - For real world applications like video
surveillance, human machine interaction, and security
systems, face recognitionisofgreatimportance. Deeplearning
based methods have shown better performance in terms of
accuracy and speed of processing in image recognition
compared to traditional machine learning methods. This
paper presents a modified architecture of the Convolution
Neural Network (CNN) by adding two operations of
normalization to two of the layers. The operation of
normalization that is normalization of the batch provided
acceleration of the network. CNN architecture was used to
extract distinctive facial characteristicsandSoftmaxclassifier
was used to classify faces within CNN's fully connected layer.
Our Face Database has shown in the experiment part that the
proposed approach has improved the performance of face
recognition with better results of recognition.
Key Words: face recognition, convolutional neural network,
softmax classifier, deep learning
1.INTRODUCTION
Face recognition is the process of recognizing a
person's face through a vision system. Because of its use in
security systems, videosurveillance,commercial areas,it has
been an important human - computer interaction tool and is
also used in social networks like Facebook. After the fast
development of artificial intelligence, face recognition has
attracted attention due toits meddlesomenatureandsinceit
is the main method of human identification when compared
with other types of biometric methods. Face recognitioncan
be easily checked in an uncontrolled environment without
the knowledge of the subject person.
As the history of face recognitionisviewed,itisseen
that it has been present in many researchpaperse.g.[1]–[6].
Traditional methods based on shallow learning have faced
challenges such as pose variation, scene lighting and facial
expression changes as in references [7]-[17]. Shallow
learning methods use only some basic image characteristics
and trust on artificial experience to extract sample
characteristics. Deep learning methods can extract more
complicated facial characteristics [18]-[27]. Deeplearningis
making crucial progress in solving issues that have formany
years restricted the artificial intelligence community's best
attempts. It has proved outstanding in disclosing complex
structures in high-dimensional data and is therefore
applicable to many science, business and government
domains. It addresses the problem of learning hierarchical
representations with a single algorithm or some algorithms
and has mainly defeated records in image recognition,
natural language processing, semantic segmentation and
many other real worldscenarios [28]-[35]. Therearevarious
deep learning approaches like Convolution Neural Network
(CNN), Deep Belief Network (DBN) [36], [37], Stacked
Autoencoder [38]. In image and face recognition, CNN is
frequently used as an algorithm. CNN is a kind of artificial
neural networks that use convolution approach to extract
characteristics from input data to increase the number of
characteristics. CNN was first proposed by LeCun and was
first used in the recognition of handwriting [39]. His
network was the source of many of the recent architectures
and an inspiration for many scientists. By publishing their
work in the ImageNet Competition, Krizhevsky, Sutskever
and Hinton achieved good results [40]. It is regarded as one
of computer vision's most dominant publications and has
shown that CNNs outperform recognition performance in
comparison with handmade methods. CNN has achieved
cutting - edge over a number of areas with computational
power from Graphical Processing Units (GPUs), including
image recognition, scene recognition and edge detection.
This paper's main contribution is to obtain a
powerful high accuracy recognition algorithm. In thispaper,
by adding Batch Normalization process after two different
layers, we developed a new CNN architecture.
In this paper, the general structure of the processof
face recognition consists of three stages. It starts with the
pre - processing stage: the conversion of color space and the
resize of images, continuing with the extraction of facial
features and then the classification of extracted feature set.
Softmax Classifier is to realize the final stage in our system,
which is classification based on the facial characteristics
extracted from CNN.
The rest of this paper is as follows organized. The
architecture of CNN is introduced in section 2. Theproposed
algorithm will be discussed in section 3. Section 4 presents
the face database used in this paper. Section 5 presents the
experimental results. Finally, in section 6, we discuss
conclusions.
2. METHODOLOGY
CNNs are a category of neural networks that have
shown to be highly effective in areas suchasfacerecognition
and classification. CNNs are a type of feed-forward, multi-
layered neural networks. CNNs consist of neurons with
learning weights. Each filter takes certain inputs, converts
and follows them with a non-linearity [41]. As shown in

Fig.1, a typical CNN architecture is shown. CNN's structure
contains layers of Convolution, pooling,RectifiedLinearUnit
(ReLU), and Fully Connected.
Fig 1: Layers of CNN
2.1 Convolution Layer
Convolution layer performs the core building block
of a Convolutionary Network that performs most of the
heavy lifting computations. Convolution layer's main
purpose is to extract features from the image-based input
data. By learning image features usingsmall squaresofinput
image, Convolution preserves the spatial relationship
between pixels. Using a set of learnable neurons, the input
image is compressed. This creates a feature map in the
output image and then feeds the feature maps to the next
convolution layer as input data.
2.2 Pooling Layer
Pooling layer reduces each activation map's
dimensionality but still has the most important information.
The images input are divided into a set of rectangles thatare
not overlapping. A non-linear operation such as average is
used to down-sample each region. This layer achievesbetter
generalization, faster convergence, robust translation and
distortion, and is usually placed between layers of
convolution.
2.3 ReLU Layer
ReLU is a non-linear operation that involves units
that use the rectifier. It is an element-wise operation which
means that it is applied per pixel, reconstituting all the
negative values by zero in the feature map. To understand
how ReLU operates, we assume that in the literature for
neural networks there is a neuron input given as x and from
that the rectifier is defined as f(x)= max(0,x).
2.4 Fully Connected Layer
The term Fully ConnectedLayer(FCL)refersto each
filter connected in the next layer in the previous layer. The
output from the layers of convolution, pooling, andReLUare
incarnation of the input image's high-level features. Theaim
of using the FCL is to use these features to classify the input
image into different classes based on the training dataset.
FCL is considered to be the final layer of pooling feeding the
features to a classifier using the activation function of
Softmax. The sum of Fully Connected Layer's output
probabilities is 1. Using the Softmax as the activation
function ensures this. The Softmax function takes an
arbitrary real-evaluated scores vector and transformsitinto
a value vector between 0 and 1 that sums up to 1.
3. THE PROPOSED ALGORITHM
The block scheme of the proposed algorithm for
CNN recognition is shown in Fig. 2. The algorithm is
performed in the following three steps:
1) Resize the input images as 16x16x1, 16x16x3,
32x32x1, 32x32x3, 64x64x1, and 64x64x1.
2) Make a CNN structure with eight layers made up of
convolutional, max pooling, convolutional, max
pooling, convolutional, max pooling, convolutional,
and convolutional layers respectively.
3) Use Softmax classifier for classification after
extracting all the features.
Fig 2: CNN Block Diagram
In picture. 3, The structure of the proposed CNN extraction
block feature is shown.
4. DATABASE
Our face database contains imagesof4 peopletaken
at different times with different lighting conditions taken
between 2/03/2019 to 8/03/19. Each individual in the
database is represented by hundred colored JPEG images
with cluttered background taken at 640 pixels resolution. In
these images, the average size of the faces is 150bis150
pixels. The pictures show frontal and/or inclined faces with
different conditions of lighting and scale. Photograph. Fig. 4
presents some face images from our face database of
different subjects [42].

5. EXPERIMENTAL RESULTS
We designed our CNN with the MatConvNet
software tool Beta23 version. The size of each image was
changed after the pre-processingstageas16x16x1,16x16x3,
32x32x1, 32x32x3, 64x64x1, and 64x64x3. 70% of the
pictures were assigned to the training set, 30% to the test
set. By making changes in image size, learning rate, batch
size, and so on, we implemented various tests. For 35
epochs, CNN was trained. The performance of the proposed
CNN was assessed based on top-1 and top-5 errors. Top-1
error rate checks whether the top class is identical to the
target label and top-5 error rate checks whether the target
label is one of your top five predictions. Table 1 shows a
brief structure of the proposed algorithm. The results in the
literature are better than those using shallow learning
techniques like in references [43-45].
Figure 5 shows the performance of the proposed
CNN architecture with respect to the top-1 error rate. As
seen from Figure 5, from the image size 64x64x3, the lowest
top-1 error rate was obtained. This result matters when it is
intended to find any subject's target label in the database.
Top-5 error rate is given in Figure 6 and the lowest ratewith
3 channels was achieved from all images.
6. CONCLUSION
This paper presents an empirical assessment of the
CNN architecture-based face recognition system. The
prominent features of the proposed algorithm is that it uses
batch normalization for the outputs of the first and final
convolution layers and higheraccuracyratesareachieved by
the network. Softmax Classifier is usedtoclassifythefacesin
a fully connected layer step. Our Face Database tested the
performance of the proposed algorithm. The resultsshowed
satisfactory rates of recognition according to literature
studies.
REFERENCES
[1] S. G. Bhele and V. H. Mankar, “A Review Paper on Face
Recognition Techniques,” Int. J. Adv. Res. Comput. Eng.
Technol., vol. 1, no. 8, pp. 2278–1323, 2012.
[2] V. Bruce and A. Young, “Understanding face recognition,”
Br. J. Psychol., vol. 77, no. 3, pp. 305–327, 1986.
[3] D. N. Parmar and B. B. Mehta, “Face Recognition Methods
& Applications,” Int. J. Comput. Technol. Appl., vol. 4, no. 1,
pp. 84–86, 2013.
[4] W. Zhao et al., “Face Recognition: A Literature Survey,”
ACM Comput. Surv., vol. 35, no. 4, pp. 399–458, 2003.

[5] K. Delac, Recent Advances in Face Recognition. 2008.
[6] A. S. Tolba, A. H. El-baz, and A. A. El-Harby, “Face
Recognition : A Literature Review,” Int. J. Signal Process.,vol.
2, no. 2, pp. 88–103, 2006.
[7] C. Geng and X. Jiang, “Face recognitionusingsiftfeatures,”
in Proceedings - International Conference on Image
Processing, ICIP, pp. 3313–3316, 2009.
[8] S. J. Wang, J. Yang, N. Zhang, and C. G. Zhou, “Tensor
Discriminant Color Space for Face Recognition,” IEEE Trans.
Image Process., vol. 20, no. 9, pp. 2490–501, 2011.
[9] S. N. Borade, R. R. Deshmukh, and S. Ramu, “Face
recognition using fusion of PCA and LDA: Borda count
approach,” in 24th Mediterranean Conference on Control
and Automation, MED 2016, pp. 1164–1167, 2016.
[10] M. A. Turk and A. P. Pentland, “Face Recognition Using
Eigenfaces,” Journal of Cognitive Neuroscience, vol. 3, no. 1.
pp. 72–86, 1991.
[11] M. O. Simón, “Improved RGB-D-T based face
recognition,” IET Biometrics, vol. 5, no. 4, pp. 297–303, Dec.
2016.
[12] O. Dniz, G. Bueno, J. Salido, and F. De La Torre, “Face
recognition usingHistogramsofOrientedGradients,”Pattern
Recognit. Lett., vol. 32, no. 12, pp. 1598–1603, 2011.
[13] J. Wright, A. Y. Yang, A. Ganesh, S. S. Sastry, and Y. Ma,
“Robust face recognition via sparse representation,” IEEE
Trans. Pattern Anal. Mach. Intell., vol. 31, no. 2, pp. 210–227,
2009.
[14] C. Zhou, L. Wang, Q. Zhang, and X. Wei,“Facerecognition
based on PCA image reconstruction and LDA,” Opt. - Int. J.
Light Electron Opt., vol. 124, no. 22, pp. 5599–5603, 2013.
[15] Z. Lei, D. Yi and S. Z. Li, “Learning Stacked Image
Descriptor for Face Recognition,” IEEE Trans. Circuits Syst.
Video Technol., vol. 26, no. 9, pp. 1685–1696, Sep. 2016.
[16] P. Sukhija, S. Behal, and P. Singh, “Face Recognition
System Using Genetic Algorithm,” in Procedia Computer
Science, vol. 85, 2016.
[17] S. Liao, A. K. Jain, and S. Z. Li, “Partial face recognition:
Alignmentfree approach,” IEEE Trans. Pattern Anal. Mach.
Intell., vol. 35, no. 5, pp. 1193–1205, 2013.
[18] Z. Zhang, P. Luo, C. C. Loy, and X. Tang, “Learning Deep
Representation for Face Alignment with Auxiliary
Attributes,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 38,
no. 5, pp. 918–930, 2016.
[19] G. B. Huang, H. Lee, and E. Learned-Miller, “Learning
hierarchical representations for face verification with
convolutional deep belief networks,” in Proceedings of the
IEEE Computer Society Conference on Computer Vision and
Pattern Recognition, pp. 2518–2525, 2012.
[20] S. Lawrence, C. L. Giles, Ah Chung Tsoi, and A. D. Back,
“Face recognition: a convolutional neural-network
approach,” IEEE Trans. Neural Networks, vol. 8, no. 1, pp.
98–113, 1997.
[21] O. M. Parkhi, A. Vedaldi, and A. Zisserman, “Deep Face
Recognition,” in Procedings of the British Machine Vision
Conference 2015, pp. 41.1- 41.12, 2015.
[22] Z. P. Fu, Y. N. ZhANG, and H. Y. Hou, “Survey of deep
learning in face recognition,” in IEEE International
Conference on Orange Technologies, ICOT 2014, pp. 5–8,
2014.
[23] X. Chen, B. Xiao, C. Wang, X. Cai, Z. Lv, and Y. Shi,
“Modular hierarchical feature learning with deep neural
networks for face verification,” Image Processing (ICIP),
2013 20th IEEEInternational Conferenceon.pp.3690–3694,
2013.
[24] Y. Sun, D. Liang, X. Wang, and X. Tang, “DeepID3: Face
Recognition withVeryDeepNeural Networks,”Cvpr,pp. 2–6,
2015.
[25] G. Hu, “When Face Recognition Meets with Deep
Learning: An Evaluation of Convolutional Neural Networks
for Face Recognition,” 2015 IEEE Int. Conf. Comput. Vis.
Work., pp. 384–392, 2015.
[26] C. Ding and D. Tao, “Robust Face Recognition via
Multimodal Deep Face Representation,” IEEE Trans.
Multimed., vol. 17, no. 11, pp. 2049– 2058, 2015.
[27] A. Bharati, R. Singh, M. Vatsa, and K. W. Bowyer,
“Detecting Facial Retouching Using Supervised Deep
Learning,” IEEE Trans. Inf. Forensics Secur., vol.11,no.9,pp.
1903–1913, 2016.
[28] M. Liang and X. Hu, “Recurrent convolutional neural
network for object recognition,” 2015 IEEE Conference on
Computer Vision and Pattern Recognition(CVPR).pp.3367–
3375, 2015.
[29] P. Pinheiro and R. Collobert, “Recurrent convolutional
neural networks for scene labeling,” Proc. 31st Int.Conf., vol.
32, no. June, pp. 82–90, 2014.
[30] W. Shen, X. Wang, Y. Wang, X. Bai, and Z. Zhang,
“DeepContour: A deep convolutional feature learned by
positive-sharing loss for contour detection,” in Proceedings
of the IEEE Computer Society Conference on Computer

Vision and Pattern Recognition, vol. 07–12, pp. 3982–3991,
June 2015.
[31] M. A. K. Mohamed, A. El-Sayed Yarub, and A. Estaitia,
“Automated Edge Detection Using Convolutional Neural
Network,” Int. J. Adv. Comput. Sci. Appl., vol.4,no.10,pp.11–
17, 2013.
[32] Dan Cireúan, “Deep Neural Networks for Pattern
Recognition.”
[33] R. Collobert, J. Weston, L. Bottou, M. Karlen, K.
Kavukcuoglu, and P. Kuksa, “Natural Language Processing
(Almost) from Scratch,” J. Mach. Learn. Res., vol. 12, pp.
2493–2537, 2011.
[34] R. Collobert and J. Weston, “A unified architecture for
natural language processing: Deep neural networks with
multitask learning,” Proc. 25th Int. Conf. Mach. Learn., pp.
160–167, 2008.
[35] E. Shelhamer, J. Long, and T. Darrell, “Fully
Convolutional Networks for Semantic Segmentation,” IEEE
Trans. Pattern Anal. Mach. Intell., vol. 39, no. 4, pp. 640–651,
2017.
[36] R. Xia, J. Deng, B. Schuller, and Y. Liu, “Modeling gender
information for emotion recognition using Denoising
autoencoder,” in ICASSP, IEEE International Conference on
Acoustics, Speech and Signal Processing – Proceedings, pp.
990–994, 2014.
[37] G. E. Hinton, S. Osindero, and Y. W. Teh, “A fast learning
algorithm for deep belief nets,”Neural Comput.,vol.18,no.7,
pp. 1527–1554, 2006.
[38] Y. Bengio, “Learning Deep Architectures for AI,” vol. 2,
no. 1, 2009.
[39] Y. LeCun, “Backpropagation AppliedtoHandwritten Zip
Code Recognition,” Neural Comput., vol. 1, no. 4, pp. 541–
551, Dec. 1989.
[40] A. Krizhevsky, I. Sutskever, and H. E. Geoffrey,
“ImageNet Classification with Deep Convolutional Neural
Networks,” Adv. Neural Inf. Process. Syst. 25, pp. 1–9, 2012.
[41] A. Uçar, Y. Demir, and C. Guzelis, “Object Recognition
and Detection with Deep Learning for Autonomous Driving
Applications,” Simulation, pp. 1-11, 2017.
[42] “Georgia Tech face database,” 10-Jan-2017 [Online].
Available: http:/www.anefian.com/research/facereco.htm.
[43] Nischal K N, Praveen Nayak M, K Manikantan, and S
Ramachandran,“FaceRecognitionusingEntropy-augmented
face isolation and Image folding as pre-processing
techniques,” 2013 Annual IEEEIndia Conference(INDICON),
2013.
[44] Katia Estabridis, “Face Recognition and Learning via
Adaptive Dictionaries,” IEEE ConferenceonTechnologies for
Homeland Security (HST), 2012.
[45] Qiong Kang and Lingling Peng, “An Extended PCA and
LDA for color face recognition,” International Conference on
Information Security and Intelligence Control (ISIC), 2012.

IRJET- Face Recognition using Machine Learning

More Related Content

What's hot (18)

Similar to IRJET- Face Recognition using Machine Learning (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Face Recognition using Machine Learning