Semantic Assisted Convolutional Neural Networks in Face Recognition

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 04 Issue: 09 | Sep -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 1257
Semantic Assisted Convolutional Neural Networks in Face Recognition
Kalyani Deepak Sonawane
Master of Engineering Student, Datta Meghe College of Engineering, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract –In today's age of automation; face recognition is
a vital component for authorization and security. The
principal aim of facial analysis is to extract valuable
information from face images, such as its position in the
image, facial characteristics, facial expressions, the person’s
gender or identity It has received substantial attention from
researchers in various fields of science such as biometrics and
computer vision. A convolutional neural network is a deep
learning algorithm that is used in object recognition. The
philosophy of a CNN is to train the network, the same way asa
human learns things. Theproposedsystem isanewframework
to efficiently and accurately match face images that are
automatically acquired under less-constrained environments.
Our framework, referred to as semantics-assisted
convolutional neural networks (SCNNs), incorporates explicit
semantic informationtoautomaticallyrecovercomprehensive
face features. The proposed system is a new framework to
efficiently and accurately match face images that are
automatically acquired under less-constrained environments.
The paper refers to semantics-assisted convolutional neural
networks (SCNNs), incorporates explicitsemantic information
to automatically recover comprehensive face features.
Key Words: Face Recognition, Artificial Neural Neuron
Architecture, Neural Network, CNN, SCNN
1. INTRODUCTION
Face recognition is one of the most relevant applications of
image analysis. Face recognition is a very challenging
research area in computer vision and pattern recognition
due to variations in facial expressions, poses and
illumination. Several emerging applications, from law
enforcement to commercial tasks, demand the industry to
develop efficient and automated face recognition systems.
It’s a true challenge to build an automated system which
equals human ability to recognize faces.Facerecognition isa
visual pattern recognition problem. Facerecognitionsystem
with the input of an arbitrary image will search in database
to output people’s identification in the input image. Face
recognition is one of the biometric methods that to have the
merits of both high accuracy and low intrusiveness. F ace
recognition is an interesting and successful application of
Pattern recognition and Image analysis. Facial images are
essential for intelligent vision-based human computer
interaction. Face processing is based on the fact that the
information about a user’s identitycanbe extractedfrom the
images and the computers can act accordingly. It is also
useful in human computer communication, computer
operated reality, database retrieval, multimedia, computer
entertainment, information security - operating system,
medical records, online banking. Biometric – personal
identification - passports,driverlicenses,automatedidentity
verification – border controls, law enforcement - video
surveillances, investigation, personal security – driver
monitoring system, home video surveillance system. A face
recognition system generally consists of four modules as
detection, alignment, feature extraction, and matching
Fig-1: Flowchart of Face Recognition system
2. NEURAL NETWORKS
Many pattern recognition problems like object recognition,
character recognition, etc. have been faced successfully by
neural networks. These systems can be used in face
detection in different ways.
The simplest definition of a neural network, more properly
referred to as an 'artificial' neural network (ANN), is
provided by the inventor of one of the first neurocomputers,
Dr. Robert Hecht-Nielsen. He defines a neural network as:
"...a computing system madeupofa numberofsimple,highly
interconnected processing elements, which process
information by their dynamic state response to external
inputs.”
Motivated right from its inception by the recognition a
machine that is designed to model the way in which the
brain performs a particular task. A massively parallel
distributed processor.

Resembles the brain in two respects:
1. Knowledge is acquired through a learning process.
2. Synaptic weights, are used to store the acquired
knowledge
Fig-2: Human brain Neuron resembles Artificial Neuron
Fig-3: Architecture of Artificial Neuron
3. CONVOLUTIONAL NEURAL NETWORKS
Different approaches have been investigated and proposed
for this task. The learning algorithms that are used in deep
learning are based on how a human learns things. A
convolutional neural network is a deep learning algorithm
that is used in object recognition. CNNs show to be a
powerful and flexible feature extraction and classification
technique which has been successfully applied in other
contexts, i.e. hand-written character recognition, and which
is very appropriate for face analysis problems.
Convolutional neural networks (CNNs) are composed of a
hierarchy of units containing a convolutional, pooling (e.g.
max or sum) and non-linear layer (e.g. ReLU max(0,x)).
In particular, unlike a regular Neural Network,thelayersofa
ConvNet have neurons arranged in 3 dimensions: width,
height, depth. (Note that the word depth here refers to the
third dimension of an activationvolume,nottothedepth ofa
full Neural Network, which can refer to the total number of
layers in a network.)
The three basic components to define a basic convolutional
network:
1. The convolutional layer
2. The Pooling layer
3. The output layer
Fig-4: Architecture of CNN
Convolutional Layer: The convolution operation extracts
different features of the input. The first convolution layer
extracts low-level features like edges, lines, and corners.
Higher-level layers extract higher-level features.
Pooling or Sub-Sampling Layer: The pooling/sub-sampling
layer reduces the resolution of the features. It makes the
features robust against noise and distortion. There are two
ways to do pooling: max pooling and average pooling. In
both cases, the input is divided into non-overlapping two-
dimensional spaces.
4. SEMANTIC ASSISTED CONVOLUTIONAL
NEURAL NETWORKS
In addition to successfully investigatingthestrengthsofCNN
for the less-constrained face recognition, our system
introduces the Semantics-Assisted CNN (SCNN)
architecture to fully exploit the discriminative information
within limited number of training samples. The approach to
the proposed system is as follows:
1. The SCNN is trained with one database and tested on
totally independent/separate databases.
2. Can also enable recovery of more comprehensive face
features from the limited training samples.
4.1 Limitations of Contemporary CNN
 To achieve superior performance using CNN based
methods, a common way is to add more layers to
make the network deeperandmorecomprehensive,

and/or devote more labeled training data because
CNN is usually trained in a supervised manner.
 Hard to afford to train such deep networks due to
the lack of enough
 The network goes deeper; the needfortrainingdata
grows accordingly computational power.
 It is difficult to acquire enough labeled training
samples.
 For instance, where the developed CNN is not very
deep (nine layers), a total of ∼200,000 face images
from more than 10,000 people were used for
training to achieve superior performance.
Therefore, we are motivated to improve the performance of
existing CNN based architecture in another way -toenhance
CNN with supervision from explicit semantic information
Fig-5: Architecture of SCNN for Face Recognition
As from above Figure, we simply add a branch, which is also
a CNN, to the existing CNN. The attached CNN is not trained
using the identity of the training data but the semantic
groups. For example, we could train CNN2 using the gender
information of the training sample, i.e., let the CNN2 be able
to estimate the gender instead of identity, and train CNN3
using the ethnicity information. After the CNNs are trained,
we can combine the output of each CNN in the wayoffeature
fusion. We refer to such extended structure of the CNN as
Semantics-Assisted CNN (SCNN for short).
4.2 Benefits of SCNN over CNN and NN
 Very helpful for identification task
 The training scheme for SCNN can reuse the same
set of training data but just labeled in another way
than the simple identities.
 The SCNN architecture and training scheme is
naturally compatible.
 This technique approach is better in performance
over other techniques due to high accuracy rate for
complex face recognition, adaptive learning as well
as better tolerance factor.
 SCNN is capable of recovering more comprehensive
features from the images and therefore achieve
superior performance.
CONCLUSION
Face recognition might be a very easytask forhumanbeings,
but it is extremely difficult to make a machine detect and
recognize human faces. In this work it hasbeenshownthatif
a facial image of a person is given then the network can able
to recognize the face of the person. The whole work is
completed through the following steps: Facial image of a
person has been collected by taking three different samples
of the person for the experiment
In this research a CNN-based face detector is used to look if
the size of the training data is of impact on the performance
of a CNN-based face detector. In particular, we proposed a
robust and more accurate framework for the face
recognition using the semantics-assisted convolutional
neural network (SCNN). By training oneormorebranchesof
CNNs with semantically information corresponding to
training data, the SCNN is capable of recovering more
comprehensive features from the images and therefore
achieve superior performance.
REFERENCES
[1] Thai Hoang Le, “Applying Artificial Neural Networks
for Face Recognition”,2011
[2] Ernst Kussul,Tetyana Baydyk,“FaceRecognitionUsing
Special Neural Networks”,2015
[3] Md. Zahangir Alom, Paheding Sidike, Vijayan K. Asari,
Tarek M. Taha, “State Preserving Extreme Learning
Machine for Face Recognition”,2015
[4] Suhas S.Satonkar, Vaibhav M.Pathak, Dr. Prakash B.
Khanale,“FaceRecognitionUsingPrincipal Component
Analysis and Artificial Neural Network of Facial
Images Datasets”,2015
[5] Zijing Zhao, Student Member, IEEE, and Ajay Kumar,
Senior Member, IEEE, “Accurate Periocular
Recognition Under Less Constrained Environment
Using Semantics-Assisted Convolutional Neural
Network”,2016

[6] Bong-Nam Kang, Yonghyun Kim, y and Daijin Kim,
“Deep Convolution Neural Network with Stacks of
Multi-scale Convolutional Layer Block using Triplet of
Faces for Face Recognition in Wild”,2016
[7] Aruni RoyChowdhury Tsung-Yu Lin Subhransu Maji
Erik Learned-Miller, “One-to-many face recognition
with bilinear CNNs” ,2016
[8] Jianxin Wu, “Introduction to Convolutional Neural
Networks”
[9] L. Nie, A. Kumar, and S. Zhan, “Periocular recognition
using unsupervised convolutional RBM feature
learning,” in Proc. 22nd Int. Conf. Pattern Recognit.
(ICPR), Stockholm, Sweden, Aug. 2014

Semantic Assisted Convolutional Neural Networks in Face Recognition

More Related Content

What's hot (17)

Similar to Semantic Assisted Convolutional Neural Networks in Face Recognition (20)

More from IRJET Journal (20)

Recently uploaded (20)

Semantic Assisted Convolutional Neural Networks in Face Recognition