FINGERPRINT BASED GENDER CLASSIFICATION USING

ARTIFICIAL NEURAL NETWORK

PROJECT REPORT
Submitted by

BINESH R B (CEA18CS015)

To

APJ Abdul Kalam Technological University

In partial fulfillment of the requirements for the award of the Degree
Of

Bachelor of Technology
In

Computer Science and Engineering

Department of Computer Science and Engineering

College of Engineering, Attingal - 695101
Trivandrum
June 2022
 Department of Computer Science and Engineering
COLLEGE OF ENGINEERING ATTINGAL
(Managed by IHRD. Govt of Kerala Undertaking)

CERTIFICATE

This is to certify that the seminar report on “FINGERPRINT BASED GENDER

CLASSIFICATION USING ARTIFICIAL NEURAL NETWORK”
submitted by BINESH R B (CEA18CS015) of eighth semester B-Tech in
Computer Science & Engineering, towards the partial fulfillment of the
requirement as a part of the curriculum for the award of the Degree of Bachelor
of Technology by APJ Abdul Kalam Technology University (KTU).

HEAD OF DEPARTMENT PROJECT GUIDE

Ms. SUMA L.S Ms. SHIJINA J SALIM
Asst Professor Asst Professor
Computer Science and Engineering Computer Science and Engineering
 ACKNOWLEDGEMENT

We thank God almighty for guiding us throughout the project. We would like to
thank all those who have contributed to the completion of the project and helped
us with valuable suggestions for improvement.

We are extremely thankful to Prof. Sunil T.T, Principal College of Engineering

Attingal for providing opportunity and the environment to the project design in
our college. Also thankful to Ms. Suma S. L, Head of the Department of
Computer Science and Engineering and our guide Ms. Shijina J Salim,
Department of Computer Science and Engineering, for providing us with best
faculties and atmosphere for the creating work guidance and encouragement.
Also, we are thankful to all our teaching and nonteaching staff in the department
and friends for extending their warm kindness and help.

Above all we would like to thank our parents, without the blessings we would not
have been able to accomplish our goal.

BINESH R B (CEA18CSS015)

i
 ABSTRACT

Fingerprint based gender classification helps to analyze the data in easy way
and help to sort out the data. Now a days fingerprint-based gender
classification is seen in civilian industrial, commercial, and unique Id of
nation as AADHAR card. As the fingerprints are unique, gender
classification helps to minimize the large data. It is utilized by using simple
scanner which is available in affordable prices. This review study introduces
a method of fingerprint based gender classification, using artificial neural
network. It will help to design and develop an efficient algorithm for the
fingerprint based intelligent gender identification system and after testing the
samples, the performance efficiency of the proposed system found to be 88%
for female and 78 % for male. Fingerprints, as one of the most widely used
biometric modalities, can be used to identify and distinguish between genders.
Gender classification is very important in reducing the time when
investigating criminal offenders and gender impersonation. In this work, we
use deep Convolutional Neural Networks (CNNs) to not only classify
fingerprints by gender, but also identify individual hands and fingers. Transfer
learning is employed to speed up the training of the CNN. The CNN achieves
an accuracy of 75.2%, 93.5%, and 76.72% for the classification of gender,
hand, and fingers, respectively. These results obtained using our publicly
available Sokoto Coventry Fingerprint Dataset (SOCOFing) serve as
benchmark classification results on this dataset.

ii
 CONTENTS

ACKNOWLEDGEMENT..……………………………………………………...i

ABSTRACT…………………………………………………………………….ii

LIST OF FIGURES…………………………………………..….……………....v

1. INTRODUCTION………………………………………………….…….….01

1.1 GENERAL BACKGROUND………………………………….………..01

1.2 OBJECTIVES…………………………………………...……..…….....03

1.3 IMPLEMENTATION OF PROJECT…………….………….………....03

2. LITERATURE SURVEY………………………………….….…………….04

3. METHODOLOGY……………………………………….……………….....07

3.1 ALGORITHM…………………………………………...……………...07

3.2 DESIGN……...………………………………………………………....08

3.2.1 FLOWCHART……………………………………...….…………...08

3.3 DATASET………………………...…………………….……...……....09

3.3.1 DATASET PREPARATION…………………...….……...………..09

3.3.2 DATA VISUALIZATION…………………………...……………..09

3.3.3 IMAGE PREPROCESSING………………………………………..09

3.4 INPUT IMAGES………………....………………………...…………...10

3.5 BLOCK DIAGRAM……………...…………………………………….10

4. SYSTEM REQUIREMENTS AND SPECIFICATIONS…………...………11

4.1 HARDWARE COMPONENTS………………………………….……..11

iii
 4.1.1 PROCESSOR……………………………………….……………....11

4.1.2 RAM……………………..………………………………………….11

4.1.3 HARDISK…………………………………………………………..11

4.2 SOFTWARE COMPONENTS………………………………………….12

4.2.1 PYTHON…..….……..………….………....………….……….……12

4.2.2 TENSOR FLOW……………………………………………………12

4.2.3 KERAS………………………………………………...……………12

5. ARCHITECTURE…………………………………………..…...…………..14

5.1.1 MODEL ARCHITECTURE……………………………………..……14

5.1.2 CONVOLUTIONAL LAYER……..……………………………...….15

5.1.3 POOLING LAYER……….…………………………………………..16

5.1.4 FULLY CONNECTED LAYER……………………….…….……….16

6. MODEL TRAINING AND VALIDATION……………………..…….…….18

7. RESULTS……………………………………………………………………20

7.1 OVERALL PERFORMANCE EVALUATION………………..……….20

7.2 OUTPUTS………………………………………………………………21

8. CONCLUSION…………………………………………………………...….22

REFERENCES…………………………………………………………………23

iv
iv
LIST OF FIGURES

Figure 3.2.1 Flowchart..............………………………………………….………08

Figure 3.4 Input Image..................................................................................…….10

Figure 3.5 Block diagram.......................…………………………………..……......................10

Figure 5.1.1 Model Architecture......…………………………………………….......................14

Figure 5.1.2 Model Summary.............………………………………......................……………15

Figure 5.2 Convolutional layer…………….......………………………..……….......................15

Figure 5.3 Pooling Layer…………………........................………....................……...………16

Figure 5.4 Fully Connected Layer…………………....………................………..17

Figure 6.1 Results of Training.........................................................................…...19

Figure 6.2 Graph shows Accuracy, Loss, Validation loss and accuracy.................19

Figure 7.1.1 Model accuracy, Validation loss and accuracy....................................20

Figure 7.1.2 Graph plot the Model Accuracy, Loss, Validation accuracy and loss....21

Figure 7.2 Output...................................................................................................21

v
 Fingerprint Based Gender Classification Using Artificial Neural Network

CHAPTER 1
INTRODUCTION

1.1 GENERAL BACKGROUND

Fingerprints have been extensively used for person identification. Fingerprints

are the unique patterns that are persistent and cannot be easily change .
Fingerprints are composed of many ridges and furrows, and they are a unique
marker of a person. No two identical fingerprints have ever been found. Features
of the fingerprints statistically differ between genders and age categories .
Personal identification is essential in security and video surveillance applications.
An individual can be recognized by various features, such as body, voice, stature,
and shape. Gender is one of the most essential features that separates between
people. Fingerprinting is considered the best technique for distinguishing
between individuals and for tracking criminals . And it has been proven by many
researchers that fingerprints can be used for gender classification, which is
helpful when short-listing suspects .

The uniqueness of the fingerprint features can reduce the gender identification
difficulties and limit the amount of time it takes to identify a suspect. These
unique features of fingerprints can be used in differentiating between individuals
by their gender and, therefore, it makes faster the identification processes of
unknown suspects. Furthermore, it can guide forensic investigators to the correct
identity of a suspect when matching the suspect’s fingerprints among the large
number of possible matches in the fingerprint databases. It is extremely time
consuming to identify and detect unknown fingerprints in a large volume of
fingerprint databases during investigation. However, by knowing the gender,
hands or fingers will grossly reduce the time it takes in identifying the culprit in

Department of Computer Science and Engineering 1 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

a large volume fingerprint database. Usually, at the initial stage of forensic

investigation, if a suspect can be deduced to be either a male or female, the
forensic investigators will limit the scope of the investigation and hence speed up
the forensic investigation.

The use of biometric fingerprints to unlock mobile phones is on the high increase
due the added security to mobile phones, yet they can be hacked with spoofed
fingers. This is an indication that a mobile phone can be unlocked irrespective of
gender. Building a model that can identify the gender is really necessary to boost
the security of the mobile phones technology.

At the border control, such systems will also help in detecting and differentiating
between genders. This is because some individuals may impersonate gender with
the intention of evading identification. Gender crime impersonation is a challenge
that needs to be solved not only through physical investigation but also using
biometric fingerprints. Crime impersonation is possible as indicated in , where
credentials like international passports, certificates and driver licenses can be
forged by deceitful criminals with technical skills despite the security features
added in them. However, biometric features like fingerprints will help reduce the
threat of counterfeiting. In this project, we propose using CNNs for gender, hand
and finger classification and report benchmark results on our publicly available
SOCOFing dataset.

Department of Computer Science and Engineering 2 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

1.2 OBJECTIVES
The main objective is to incorporate a system which is able to identify the
gender of the individual.

To design as a tool to help the forensic investigation.

To identify the gender we use Convolutional Neural Networks (CNN’s or

ConvNet) to analyze the fingerprint.

When the system gets an input image of the fingerprint, the system will output
the gender as 1 or 0 which determines male or female.

1.3 IMPLEMENTATION OF PROJECT

The system mainly consists of three phases

Preprocessing the Dataset.

Building a model for gender identification.

Detection of gender using the model which is built.

Department of Computer Science and Engineering 3 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

CHAPTER 2
LITERATURE SURVEY

Different studies on gender determination carried out so far are based on the
three-dimensional domain scrutiny of ridges. Previous studies have used ridge
associated parameters for gender identification, which include: fingerprint ridge
tally, thickness, thickness to valley proportion, width, finger impression designs
and design styles. Some experimental results presented in many studies show that
females reveal a greater ridge thickness due to better and clearer epidermal ridge
details compared to males. Nithin et al present a study which aimed at
determining the gender based on finger ridge count within a distinct region of
Southern India. A sample of 550 rolled-finger fingerprints was taken, in which
275 were from men and 275 from women, all of them in the age limits of 18 - 65
years. The outcomes of the research indicate that female have an ominously
greater ridge count than their male counterpart. The application of Bayes’ theory
in this study suggests that a fingerprint possessing ridge density less than 13
ridges per 25 mm2 is most likely to be of male origin. Likewise, a fingerprint
having a ridge count greater than 14 ridges per 25 mm2 is most likely to be of
female origin.

Another study on different techniques for gender identification using

fingerprints. The study proposed different ridge related approaches used for
gender identification, like unique finger impression edge check, thickness, width
and unique finger impression patterns. The study identifies the frequency domain
approaches as the most efficient method, because of its flexibility and less
computation time required when compared with the spatial domain approaches.

Gnanasivam, P. and Muttan, D.S., 2012 conducted a research on fingerprint

gender classification using wavelet transforms and Singular Value
Decomposition (SVD) methods. A sample of 3570 fingerprints was taken, in

Department of Computer Science and Engineering 4 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

which 1980 fingerprints all from male and 1590 fingerprints all from female.
They proposed a new method for classifying gender based on fingerprint images,
by using discrete wavelet transforms and Singular Value Decomposition. The
techniques considered the spatial features of the singular value decomposition,
which includes the internal structure of the fingerprint images and the frequency
features of the wavelet domain, and which improved performance in gender
classification. The result indicates that the female fingerprints show significantly
higher classification rates than those of males. Gender classification for subjects
male tested attained 91.67% accuracy rate, and 84.69% for females, respectively.

Classification of gender from fingerprints was also proposed by, Tadross, R. and
Jantz, R., 2006, which recognize the gender of a criminal and help in decreasing
the suspect search time. Their research analyzed a dataset of 10 fingerprint images
for 2200 people of different gender and age using feature extraction techniques.
Neural networks, Fuzzy C-means and linear discriminant analysis were used, and
classification results of 88.5%, 80.39%, and 86.5% were obtained, respectively.
Furthermore, an implementation of a novel method to classify gender was
presented by Gupta, S. and Rao, A.P in 2014 using two combined methods of
wavelet transformation to extract fingerprint features and pass the output to back-
propagation neural network algorithms for the final gender classification. The
experimentation was performed using a fingerprint database of 275 male and 275
female fingerprints, and obtained a classification accuracy of 91.45%.

CNN based models have proven to be robust in image classification tasks. A

Previous work has demonstrated that CNNs can be used for fingerprint
classification and produce remarkable results. Therefore, detailed classification
of fingerprint using CNNs will help provide a more effective method for gender,
hand and finger classifications. Previous methods for fingerprint gender
classification used hard coded features of the fingerprints for gender
classification. With a CNN, the model learns the features to distinguish between

Department of Computer Science and Engineering 5 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

genders, which makes it better, as it takes into consideration every bit of the
fingerprint image information. This fingerprint image information is critical for
distinguishing between fingerprint images, as the patterns of the fingerprint
images retain the required information for the classification.

Department of Computer Science and Engineering 6 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

CHAPTER 3
METHODOLOGY
The proposed method consists of a pre trained CNN model having two
convolutional layers after each layer a maxpooling layer will be added with ‘relu’
activation. The algorithm for fingerprint based gender classification is as follows:

3.1 ALGORITHM

Input : Finger print images

Output : Gender prediction
1. For each image in the dataset do
2. Resize the gray scale image into 96 x 96
3. Load the labels and extracted features into a numpy array
4. Reshape image features and shuffle the train and test data
5. End
6. For building the CNN model do
7. Add 2 alternating pairs of convolutional and maxpooling layers with 32
Filters
8. Insert a flatten layer to network classifier
9. Add a dense layer with 128 outputs
10. Add the final dense layer with 2 outputs
11. End
12. Split the data and train the model

Department of Computer Science and Engineering 7 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

3.2 DESIGN

3.2.1 FLOWCHART

PREDICT THE
FINGERPRINT
GIVEN IS MALE OR
FEMALE

Figure 3.2.1 Flowchart

Department of Computer Science and Engineering 8 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

3.3 DATASET
3.3.1 DATASET PREPARATION
The dataset was a source from Kaggle, which is a gender classification
competition website based on fingerprints. This dataset has a total of 6000 images
data that included test data and training data by classifying two categories, male
and female. The dataset used has an image size of 96x103 pixels. This size image
needs to be reduced to facilitate the training process. Before entering the training
data, preprocessing data is performed, which is resizing the fingerprint image into
96x96 pixels.

3.3.2 DATA VISUALIZATION

Data visualization is the process of transforming abstract data to meaningful

representation using knowledge communication and insight discovery through
encoding. It is helpful to study a particular pattern in the dataset. It has two main
directory — Real and Altered directory. The altered directory is further divided
into Altered-Easy, Altered-Hard and Altered-Medium. The Altered directory
contains synthetically altered versions of fingerprint images with three different
levels of alteration for obliteration, central rotation, and z-cut. The Real directory
contains Real human fingerprints (No alteration). The labels for Altered image
follows this format “100__M_Left_thumb_finger_CR.BMP”. The labels for Real
images looks like this “100__M_Left_thumb_finger.BMP.

3.3 .3 IMAGE PREPROCESSING

All the image in the specified path(path), reads and convert them to gray-scale
color(i.e black and white) then returns an array of pixel values of those images
to img_array. Our intended image size was set img_size = 96.
img_resize contains array values that have been resized based on img_size. All
images will therefore have the same size(96x96). All labels and resized image
array values were then appended to a list variable data. img contains image

Department of Computer Science and Engineering 9 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

arrays. Values in img were converted into array once more before it was
reshaped. The pixel values of images ranges from 0 to 255 by
dividing train_data by 255.0, The pixel values will be scaled down to a range
between 0 and 1

3.4 INPUT IMAGES

Figure 3.4 Input Images

3.5 BLOCK DIAGRAM

(Male or
Female)

Figure 3.5 Block Diagram

Department of Computer Science and Engineering 10 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

CHAPTER 4
SYSTEM REQUIREMENTS AND SPECIFICATION

4.1 HARDWARE COMPONENTS

4.1.1 PROCESSOR
Every computer contains at least one processor, also known as a CPU or central
processing unit. Here CPU is probably made by Intel or AMD. Processor is 1
gigahertz (GHz) or faster processor or System on a Chip (SoC). Main processor
or just processor, is the electronic circuitry that executes instructions comprising
a computer program. The CPU performs basic arithmetic, logic, controlling, and
input/output (I/O) operations specified by the instructions in the program.

4.1.2 RAM

Random-access memory (RAM) is a form of computer memory that can be read

and changed in any order, typically used to store working data and machine code.
A random access memory device allows data items to read or write in almost the
same amount of time irrespective of the physical location of data inside the
memory, in contrast with other direct access data storage media (such as hard
disks, CD-RWs, DVD-RWs and the older magnetic tapes and drum memory),
where the time required to read and write data items varies significantly
depending on their physical locations on the recording medium, due to
mechanical limitations such as media rotation speeds and arm movement. RAM
is 1 gigabyte (GB) for 32- bit or 2 GB for 64-bit.

4.1.3 HARD DISK

Hard disk, also called hard disk drive or hard drive, magnetic storage medium for
a computer. Hard disks are flat circular plates made of aluminum or glass and
coated with a magnetic material. Hard disks for personal computers can store

Department of Computer Science and Engineering 11 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

terabytes (trillions of bytes) of information. Hard disk space: 16 GB for 32-bit OS

or 20 GB for 64-bit OS.

4.2 SOFTWARE COMPONENTS

4.2.1 PYTHON
Python is a general-purpose interpreted, interactive, object-oriented, and high-
level programming language. It was created by Guido van Rossum during 1985-
1990. Like Perl, Python source code is also available under the GNU General
Public License (GPL). Python is ahigh-level, interpreted, interactive and object-
oriented scripting language. Python is designed to be highly readable. It uses
English keywords frequently where as other languages use punctuation, and it has
fewer syntactical constructions than other languages. Python is a must for
students and working professionals to become a great Software Engineer
specially when they are working in Web Development Domain.

4.2.2 TENSOR FLOW

TensorFlow is an end-to-end opensource platform for machine learning. It has a

comprehensive, flexible ecosystem of tools, libraries and community resources
that lets researchers push the state-of-the-art in ML and developers easily build
and deploy ML powered applications. Build and train ML models easily using
intuitive high-level APIs like Keras with eager execution, which makes for
immediate model iteration and easy debugging. Easily train and deploy models
in the cloud in the browser, or on-device no matter what language you use. A
simple and flexible architecture to take new ideas from concept to code, to state-
of-the-art models, and to publication faster.

4.2.3 KERAS

Keras is a deep learning API written in Python, running on top of the machine
learning platform TensorFlow. It was developed with a focus on enabling fast
experimentation. Keras is one of the leading high-level neural networks APIs. It

Department of Computer Science and Engineering 12 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

is written in Python and supports multiple back-end neural network computation

engines. Keras was created to be user friendly, modular, easy to extend, and to
work with Python. Neural layers, cost functions, optimizers, initialization
schemes, activation functions, and regularization schemes are all standalone
modules that you can combine to create new models. New modules are simple to
add, as new classes and functions. Models are defined in Python code, not
separate model configuration files.

Department of Computer Science and Engineering 13 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

CHAPTER 5
ARCHITECTURE

Convolutional Neural Network is including in the type of deep learning because

of the depth of the network. Convolutional neural network is a convolution
operation that combines several layers of processing, using several elements that
operate in parallel and are inspired by the biological nervous system. On
convolution neural network each neuron is presented in two-dimensional form,
so this method is suitable for processing with input in the form of images.
Convolutional neural network has three layers, that is convolution layer, pooling
layer, and fully connected layer. CNN model having two convolutional layers
after each layer a maxpooling layer will be added with ‘relu’ activation. Each of
the convolutional layer(Conv2D) consist of 32 filters each of size 3 (i.e 3 x 3).
Maxpooling layers(MaxPooling2D) have a pool size of 2. There is one hidden
layer(Dense) with 128 units, activation function is ‘relu’.Then a flatten layer, an
hidden dense layer then an output layer. At the end there is a network with an
output Dense layer of size 2 (class number) and a softmax activation .

5.1 MODEL ARCHITECTURE

Figure 5.1.1 Model Architecture

Department of Computer Science and Engineering 14 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

5.1.2 Model Summary

5.2 CONVOLUTIONAL LAYER

Convolution layer is a process of extracting an object from the input image using
a filter. This filter contains weights used to detect the character of an object by
multiplication between two metrices. The first matrix is an input image, and the
second is a filter producing a feature map. Each of the convolutional
layer(Conv2D) consist of 32 filters each of size 3 (i.e 3 x 3).

Figure 5.2 Convolutional Layer

Department of Computer Science and Engineering 15 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

5.3 POOLING LAYER

At this layer, the size of a data image from the convolution process will be
reducing. The pooling layer consists of filters of a specific size and stride then
shifts throughout the feature map area. For most Convolutional Neural Network
architectures, the pooling method used is max pooling. Max pooling divides the
convolution output layer into several gird, and then each filter shift will take the
most considerable value from each grid. A small portion of the image size
produced int his process is useful for reducing the dimensions of the data so that
it can reduce the number of parameters in the next step. The output of the pooling
layer is a feature map. CNN model having two convolutional layers after each
layer a maxpooling layer will be added with ‘relu’ activation. Maxpooling
layers(MaxPooling2D) have a pool size of 2.

Figure 5.3 Pooling Layer

5.4 FULLY CONNECTED LAYER

The feature map, which is output of pooling layer, is still in the form of a
multidimensional array. This layer reshapes the feature map and generates n-
dimensional vectors where n is the number of output classes the program must
select. The result is a feature map in the form of a one-dimensional vector. There
is only one hidden layer(Dense) with 128 units, activation function is ‘relu’.
At end there is a network with an output Dense layer of size 2 (class number)
and a softmax activation.

Department of Computer Science and Engineering 16 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

Figure 5.4 Fully Connected Layer

Department of Computer Science and Engineering 17 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

CHAPTER 6
MODEL TRAINING AND VALIDATION

Model training is the phase in the data science development lifecycle where
practitioners try to fit the best combination of weights and bias to a machine
learning algorithm to minimize a loss function over the prediction range. The
purpose of model training is to build the best mathematical representation of the
relationship between data features and a target label. Model training is the key
step in machine learning that results in a model ready to be validated, tested, and
deployed. The performance of the model determines the quality of the
applications that are built using it.

An epoch is a term used in machine learning and indicates the number of passes
of the entire training dataset the machine learning algorithm has completed.
Datasets are usually grouped into batches (especially when the amount of data is
very large). Some people use the term iteration loosely and refer to putting one
batch through the model as an iteration. We set epoch equal to 30. The batch size
is a hyperparameter that defines the number of samples to work through before
updating the internal model parameters. Then we fitted the model using Model.fit.
Model fitting is a measure of how well a machine learning model generalizes to
similar data to that on which it was trained. A model that is well-fitted produces
more accurate outcomes. A model that is overfitted matches the data too closely.
A model that is underfitted doesn’t match closely enough.

The model is trained it gives out values for train loss and accuracy and also
validation loss and accuracy per epoch. We successfully trained our model with
a training accuracy of 99 percent and val accuracy of 98 percent.

Department of Computer Science and Engineering 18 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

Figure 6.1 Results of Training

Figure 6.2 Graph shows Accuracy, Loss, Validation loss and accuracy

Department of Computer Science and Engineering 19 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

CHAPTER 7
RESULTS
Python programming language was used to train the proposed deep transfer
learning models. All experiments were performed on Microsoft Visual Studio
code with Windows 11 operating system and with the help of online cloud service
Google Colab and Kaggle for free.

Dataset is taken from Kaggle and it consists of roughly 55000 images. After the
preparation of dataset, we resize the images into 96×96. The image is of greyscale
format. Then we divide the dataset into training and validation.Then we train the
model with Adaptive Moment Estimation (ADAM) optimizer. Batch size and
Number of epochs set to 128 and 30 respectively for all experiments.

7.1 OVERALL PERFORMANCE EVALUATION

For evaluating the performance of proposed deep learning classifiers, we plot

model accuracy and loss, find confusion matrix and classification report including
precision recall and f1 score of dataset used for training and testing.

Figure 7.1.1 Model accuracy, Validation loss and accuracy

Department of Computer Science and Engineering 20 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

Figure 7.1.2 Graph plot the Model Accuracy, Loss, Validation accuracy and loss

7.2 OUTPUTS

Figure 7.2 Output

Department of Computer Science and Engineering 21 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

CHAPTER 8
CONCLUSION
Fingerprint gender classification helps in classifying and differentiating between
genders. We proposed using CNNs and transfer learning in this paper for gender,
hands and fingers classifications. Our proposed method achieves promising
results. The results of the classification obtained serve as benchmark results for
our publicly available SOCOFing corpus. To facilitate the training and benefit
from transferable features learned by a model in a different domain, we employed
TLDA and ResNets for fingerprint classification. The employed CNN model
presents a novel method for hands and finger classification, with an overall
classification accuracy rate of 75.52% in gender, 93.5% in hand and 76.72% in
finger classification. The usage of this technique can be considered as a main
method to be used in order to reduce the suspect search list and time by forensic
investigators, as fingerprints remain the most reliable option for personal
identification accepted in the courts of law. In future work, we will extend the
method by using some other publicly available fingerprint datasets, with less
background content as well as centred fingerprint images. We will also attempt
to improve the classification of gender and fingers by introducing the use of
minutiae points in the fully connected layer along with the fingerprint features
extracted by the convolutional layers. Moreover, we will explore whether other
factors such as fingerprint thickness or valley thickness can assist in improving
the classification performance of deep CNN models. More datasets will also be
considered as CNNs need high volume of data to learn from. We will also explore
the possibility of skin grafting, where individuals cut one part of their finger skin
and place it onto another: this may be by switching the grafted skin from a male
to female skin, or from left hand to right hand, or even switching between the
fingers.

Department of Computer Science and Engineering 22 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

REFERENCES
1. D. I. Marbun, D. Ilmu, K. Forensik, F. K. Usu, and R. H. A. Malik, “Penentuan Jenis Kelamin
Berdasarkan Kerapatan Alur Sidik Jari,” Maj. Kedokt. Nusant. J. Med. Sch., vol. 51, no. 1, pp.
6–9, 2019.

2. D. Scarlet, “What is biometrics? 10 physical and behavior identifiers that can be used for
authentication,” Journal of Chemical Information and Modeling, 2013. [Online]. Available:
https://www.csoonline.com/article/3339565/what-is-biometrics-and-whycollecting-biometric-
data-is-risky.html. [Accessed: 29-Feb-2020].

3. M. K. Shinde and S. A. Annadate, “Analysis of fingerprint image for gender classification

or identification: Using wavelet transform and singular value decomposition,” Proc. - 1st Int.
Conf. Comput. Commun. Control Autom. ICCUBEA 2015, pp. 650–654, 2015, doi:
10.1109/ICCUBEA.2015.133.

4. A. Santoso and G. Ariyanto, “Implementasi Deep Learning Berbasis Tensorflow,” J. Emit.,

vol. 18, no. 01, pp. 22–27, 2018.

5. X. Cao, L. Xu, D. Meng, Q. Zhao, and Z. Xu, “Integration of 3-dimensional discrete wavelet
transform and Markov random field for hyperspectral image classification,” Neurocomputing,
vol. 226, no. July, pp. 90–100, 2017, doi: 10.1016/j.neucom.2016.11.034.

6. Z. Li, B. Niu, F. Peng, G. Li, Z. Yang, and J. Wu, “Classification of Peanut Images Based
on Multi-features and SVM,” IFACPapersOnLine, vol. 51, no. 17, pp. 726–731, 2018, doi:
10.1016/j.ifacol.2018.08.110.

7. Z. Zhang, “Semi-supervised hyperspectral image classification algorithm based on graph

embedding and discriminative spatial information,” Microprocess. Microsyst., vol. 75, p.
103070, 2020, doi: 10.1016/j.micpro.2020.103070.

8. J. Zhao, Q. Hu, G. Liu, X. Ma, F. Chen, and M. M. Hassan, “AFA: Adversarial fingerprinting
authentication for deep neural networks,” Comput. Commun., vol. 150, no. December 2019,
pp. 488–497, 2020, doi: 10.1016/j.comcom.2019.12.016.

Department of Computer Science and Engineering 23 College of Engineering Attingal

 Fingerprint Based Gender Classification Using Artificial Neural Network

9. I. G. Y. B. and A. Courville, “Deep Learning Ian,” Foreign Aff., vol. 91, no. 5, pp. 1689–
1699, 2012, doi: 10.1017/CBO9781107415324.004.

10. M. Castelluccio, G. Poggi, C. Sansone, and L. Verdoliva, “Land Use Classification in

Remote Sensing Images by Convolutional Neural Networks,” no. September, 2015.

11. Y. Chen, X. Zhao, and X. Jia, “Spectral-Spatial Classification of Hyperspectral Data Based
on Deep Belief Network,” IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens., vol. 8, no. 6, pp.
2381–2392, 2015, doi: 10.1109/JSTARS.2015.2388577.

12. “APPLICATION OF DEEP-LEARNING ALGORITHMS TO MSTAR DATA Haipeng

Wang , Member , IEEE , Sizhe Chen , Student Member , IEEE , Feng Xu , Senior Member ,
IEEE and Ya-Qiu Jin , Fellow , IEEE Key Laboratory for Information Science of
Electromagnetic Waves ( MoE ),” IEEE Int. Geosci. Remote Sens. Symp. 2015, pp. 3743–
3745, 2015.

13. R. C. Hidayatullah and S. Violina, “Convolutional Neural Network Architecture and Data
Augmentation for Pneumonia Classification from Chest X-Rays Images,” vol. 5, no. 2, pp.
158–164, 2020.

14. W. S. Jeon and S. Y. Rhee, “Fingerprint pattern classification using convolution neural
network,” Int. J. Fuzzy Log. Intell. Syst., vol. 17, no. 3, pp. 170–176, 2017, doi:
10.5391/IJFIS.2017.17.3.170

Department of Computer Science and Engineering 24 College of Engineering Attingal