Bird Species Classification using Deep Learning Literature Survey

CHAPTER 1

INTRODUCTION

1.1 Introduction
The everyday pace of life tends to be fast and frantic and involves extramural activities. Bird
watching is a recreational activity that can provide relaxation in daily life and promote
resilience to face daily challenges. It can also offer health beets and happiness derived from
enjoying nature. Numerous people visit bird sanctuaries to glance at the various bird species
or to praise their elegant and beautiful feathers while barely recognizing the differences
between bird species and their features. Understanding such differences between species can
enhance our knowledge of exotic birds as well as their ecosystems and biodiversity.

However, because of observer constraints such as location, distance, and equipment,

identifying birds with the naked eye is based on basic characteristic features, and appropriate
classification based on distinct features is often seen as tedious. In the past, computer vision
and its subcategory of recognition, which use techniques such as machine learning, have been
extensively researched to delineate the specific features of objects, including vegetables and
fruits, landmarks, clothing, cars, plants, and birds, within a particular cluster of scenes.
However, considerable room for improvement remains in the accuracy and feasibility of bird
feature extraction techniques. Detection of object parts is challenging because of complex
variations or similar subordinate categories and fringes of objects. Intra class and interclass
variation in the silhouettes and appearances of birds is difficult to identify correctly because
certain features are shared among species.

To classify the aesthetics of birds in their natural habitats, this study developed a
method using a Convolutional neural network (CNN) to extract information from bird images
captured previously or in real time by identifying local features. First, raw input data of
myriad semantic parts of a bird were gathered and localized. Second, the feature vectors of
each generic part were detected and lettered based on shape, size, and colour. Third, a CNN
model was trained with the bird pictures in a graphics processing unit (GPU) for feature
vector extraction with consideration of the aforementioned characteristics, and subsequently
the classified, trained data were stored on a server to identify a target object.

Dept. of ISE,Vemana IT Page 1 2019-2020

Bird Species Classification using Deep Learning Literature Survey

CHAPTER 2

LITERATURE SURVEY
[1] Likeability of Garden Birds: Importance of Species Knowledge & Richness in
Connecting People to Nature.IEEE 2015
This study provides evidence that the well-being benefits that people receive from interacting
with the birds in their garden is dependent on their familiarity with different species, and that
these benefits are enhanced by increased species richness. Attention should be given to
strategies that focus on increasing the diversity of songbirds within urban green spaces, along
with increasing the ability of recreational green space users to recognise different components
of the natural environment. Combined these approaches will enhance both the well-being
benefits that people receive from interacting with nature and the biological complexity of
urban green spaces. Bird feeders provide a powerful tool for people to engage with the natural
world in their own garden, and so act as an important stepping-stone for a wider
connectedness to nature. People with a greater connection to nature are more likely to be
aware of, and support, conservation issues in the wider landscape So feeding birds can be
seen as an important tool in reconnecting people to the natural world, so helping to mediate
the extinction of experience.

Advantages:

 Bird feeders provide a powerful tool for people to engage with the natural world in their
own garden.

 It act as an important stepping-stone for a wider connectedness to nature.

Disadvantages:

 Birds face several risks when they assemble in large groups.

[2] Image Net Large Scale Visual Recognition Challenge,IEEE 2016

In this paper we described the large-scale data collection process of ILSVRC, provided a
summary of the most successful algorithms on this data, and analysed the success and failure

Dept. of ISE,Vemana IT Page 2 2019-2020

Bird Species Classification using Deep Learning Literature Survey

modes of these algorithms. In this section we discuss some of the key lessons we learned over
the years of ILSVRC, strive to address the key criticisms of the datasets and the challenges
we encountered over the years, and conclude by looking forward into the future.

Advantages:

 It works to help protect the environment.

Disadvantages:

 There is an inadequate focus on recovery.

[3] Object-Part Attention Model for Fine-Grained Image Classification,IEEE 2017

In this paper, the OPAM approach has been proposed for weakly supervised fine-grained
image classification, which jointly integrates two level attention models: object-level
localizes objects of images, and part-level selects discriminative parts of objects. The two
level attentions jointly improve the multi-view and multi-scale feature learning and enhance
their mutual promotion. Besides, part selection is driven by the object-part spatial constraint
model, which combines two spatial constraints: object spatial constraint ensures the high
representativeness of selected parts, and part spatial constraint eliminates redundancy and
enhances discrimination of selected parts. Combination of the two spatial constraints
promotes the subtle and local discrimination localization. Importantly, our OPAM avoids the
heavy labour consumption of labelling to march toward practical application. Comprehensive
experimental results show the effectiveness of our OPAM approach compared with more than
ten state-of-the-art methods on four widely-used datasets. The future work lies in two aspects:
First, we will focus on learning better fine-grained representation via more effective and
precise part localization methods. Second, we will also attempt to apply semi-supervised
learning into our work to make full use of large amounts of web data. Both of them will be
employed to further improve the fine-grained image classification performance.

Advantages:

 It provides more support for endangered species .

Dept. of ISE,Vemana IT Page 3 2019-2020

Bird Species Classification using Deep Learning Literature Survey

Disadvantages:

 Implementation of endangered species Act is chronically under-funded.

[4] Landmark Classification with Hierarchical Multi Modal Exemplar Feature ,IEEE
2018

Effective landmark classification is fundamental for many georeferenced image search and
analytics applications. One of the most challenging problems in landmark classification is how
to model and characterize landmark image, which could have high visual diversity and
complexity. In this paper, we present a novel image signature scheme called HMME to
effectively 992 IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 17, NO. 7, JULY 2015
characterize landmark images. Also, an effective exemplar selection approach is proposed to
mine hierarchical exemplars in multiple modalities from large amount of candidate images and
regions. An effective feature generation framework based on region coding is developed to
generate feature dimensions. Based on hierarchical multi-modal exemplars, HMME can
effectively represent diverse visual contents of landmark images. Further, with region based
semantic coding, HMME can integrate heterogeneous discriminative information from
multiple modalities and various spatial levels, and enjoy superior robustness against visual
variance of landmark images. Comparative experiments on real world landmark datasets
demonstrate the effectiveness of HMME compared with several state-of-the-art techniques.
Superior performance of HMME illustrates its greatest potentials for being applied to a wide
range of real world visual landmark retrieval and mining applications.

Advantages:

 There are several economic benefits that occur through the protection of wildlife.

Disadvantages:

 The endangered species Act has minimal flexibility.

[5] Object Detection with Deep Learning: A Review,IEEE 2019

Due to its powerful learning ability and advantages in dealing with occlusion, scale
transformation and background switches, deep learning based object detection has been a
research hotspot in recent years. This paper provides a detailed review on deep learning

Dept. of ISE,Vemana IT Page 4 2019-2020

Bird Species Classification using Deep Learning Literature Survey

based object detection frameworks which handle different sub-problems, such as occlusion,
clutter and low resolution, with different degrees of modifications on R-CNN. The review
starts on generic object detection pipelines which provide base architectures for other
related tasks. Then, three other common tasks, namely salient object detection, face
detection and pedestrian detection, are also briefly reviewed. Finally, we propose several
promising future directions to gain a thorough understanding of the object detection
landscape. This review is also meaningful for the developments in neural networks and
related learning systems, which provides valuable insights and guidelines for future
progress.

Advantages:

 Bird-watching provides stress relief for humans.

Disadvantages:

 Lack of certified reference materials.

Dept. of ISE,Vemana IT Page 5 2019-2020

Bird Species Classification using Deep Learning Literature Survey

CHAPTER 3

OBJECTIVE

3.1 Problem Statement

The general idea of Bird species classification is extracting features from Bird landmarks and
then compare to other bird images by matching those features. Basically bird identification is
done visually or acoustically. The main visual components comprise of bird’s shape, its wings,
size, pose, colour, etc. However, while considering the parameters time of year must be taken
into consideration because of bird’s wings changes according to their growth. The acoustics
components comprise the songs and call that birds make. The marks that distinguish one bird
from another are also useful, such as breast spots, wing bars which are described as thin lines
along the wings, eye rings, crowns, eyebrows. The shape of the beak is often an important
aspect as a bird can recognized uniquely.

3.2 Objectives
Our main objective is to build a deep learning model which will use Deep Learning approach
for the Bird Species Classification.

Deep Learning: Deep learning is a subset of machine learning in artificial intelligence (AI)
that has networks capable of learning unsupervised from data that is unstructured or unlabeled.
Also known as deep neural learning or deep neural network.

Dept. of ISE,Vemana IT Page 6 2019-2020

Bird Species Classification using Deep Learning Literature Survey

CHAPTER 4

SYSTEM REQUIREMENTS

System Requirement Specification or Software requirements specification (SRS) is a

document that describes the features and behavior of a system or software application
and captures the complete description about how the system is expected to perform. It
is typically signed off at the tip of needs engineering part. Framework Requirement
Specification (SRS) is a focal report, which outlines the foundation of the item headway
handle. It records the necessities of a structure and in addition has a delineation of its
noteworthy highlight. A SRS is basically an affiliation's seeing (in making) of a customer
or potential client's edge work necessities and conditions at a particular point in time (for
the most part) before any veritable design or change work. It's a two-way insurance
approach that ensures that both the client and the affiliation understand exchange's
necessities from that perspective at a given point in time. Every software package has
certain specific goals and serves specific functions. Each goal and
purpose interprets a method or many processes that the software package aims to
resolve or to automatize. To deliver the proper merchandise, users must always outline
the software package from the start.

System requirement specification or SRS frameworks software development,

documents every operation and dictates how software should behave, it can be as detailed
as what a button should do and should be as complete and correct as possible. The
purpose of a specification document is to describe the behavior as well as the different
functionalities of an application or software in a specific environment.

4.1 FUNCTIONAL REQUIREMENTS

A function of software system is defined in functional requirement and the behavior of the
system is evaluated when presented with specific inputs or conditions which may include
calculations, data manipulation and processing and other specific functionality. The
functional requirements of the project are one of the most important aspects in terms of entire
mechanism of modules.

Dept. of ISE,Vemana IT Page 7 2019-2020

Bird Species Classification using Deep Learning Literature Survey

Once our model is built then it should be able to classify the Birds image.

4.2 NON-FUNCTIONAL REQUIREMENTS

Non-functional requirements describe how a system must behave and establish constraints of
its functionality. This type of requirements is also known as the system’s quality attributes.
Attributes such as performance, security, usability, compatibility are not the feature of the
system, they are a required characteristic. They are "developing" properties that emerge from
the whole arrangement and hence we can't compose a particular line of code to execute them.
Any attributes required by the customer are described by the specification. We must include
only those requirements that are appropriate for our project.

Figure 4: Non-Functional Requirements

Non-Functional Requirements are as follows:

• Reliability
The structure must be reliable and strong in giving the functionalities. The movements
must be made unmistakable by the structure when a customer has revealed a couple of

Dept. of ISE,Vemana IT Page 8 2019-2020

Bird Species Classification using Deep Learning Literature Survey

enhancements. The progressions made by the Programmer must be Project pioneer

and in addition the Test designer.

• Maintainability
The system watching and upkeep should be fundamental and focus in its approach.
There should not be an excess of occupations running on diverse machines such that it
gets hard to screen whether the employments are running without lapses.

• Performance
The framework will be utilized by numerous representatives all the while. Since the
system will be encouraged on a single web server with a lone database server outside
of anyone's ability to see, execution transforms into a significant concern. The
structure should not capitulate when various customers would use everything the
while. It should allow brisk accessibility to each and every piece of its customers. For
instance, if two test specialists are all the while attempting to report the vicinity of a
bug, then there ought not to be any irregularity at the same time.

• Portability
The framework should to be effectively versatile to another framework. This is
obliged when the web server, which s facilitating the framework gets adhered because
of a few issues, which requires the framework to be taken to another framework.

• Scalability
The framework should be sufficiently adaptable to include new functionalities at a
later stage. There should be a run of the mill channel, which can oblige the new
functionalities.

• Flexibility
Flexibility is the capacity of a framework to adjust to changing situations and
circumstances, and to adapt to changes to business approaches and rules. An adaptable
framework is one that is anything but difficult to reconfigure or adjust because of
diverse client and framework prerequisites. The deliberate division of concerns

Dept. of ISE,Vemana IT Page 9 2019-2020

Bird Species Classification using Deep Learning Literature Survey

between the trough and motor parts helps adaptability as just a little bit of the
framework is influenced when strategies or principles change.

4.3 Hardware Requirements:

System : Pentium IV 2.4 GHz.
Hard Disk : 500 GB.
RAM : 4 GB

4.4 Software Requirements:

Operating System : Windows 7 / 8 / 10

Coding Language : Python
Software : Anaconda
IDE : Jupyter Notebook

Dept. of ISE,Vemana IT Page 10 2019-2020

Bird Species Classification using Deep Learning Literature Survey

CHAPTER 5

DESIGN METHODOLOGY
It will cover the details explanation of methodology that is being used to make this project
complete and working well. Many methodology or findings from this field mainly generated
into journal for others to take advantages and improve as upcoming studies. The method is
use to achieve the objective of the project that will accomplish a perfect result. In order to
evaluate this project, the methodology based on System Development Life Cycle (SDLC),
generally three major step, which is planning, implementing and analysis.

Figure 5.0: Software Development Life Cycle

Figure 5.0.1: Steps of Methodology

Dept. of ISE,Vemana IT Page 11 2019-2020

Bird Species Classification using Deep Learning Literature Survey

Planning:

To identify all the information and requirement such as hardware and software, planning must
be done in the proper manner. The planning phase has two main elements namely data
collection and the requirements of hardware and software.

Data collection:

Machine learning needs two things to work, data (lots of it) and models. When acquiring the
data, be sure to have enough features (aspect of data that can help for a prediction, like the
surface of the house to predict its price) populated to train correctly your learning model. In
general, the more data you have the better so make to come with enough rows.

The primary data collected from the online sources remains in the raw form of statements,
digits and qualitative terms. The raw data contains error, omissions and inconsistencies. It
requires corrections after careful scrutinizing the completed questionnaires.

The following steps are involved in the processing of primary data. A huge volume of raw
data collected through field survey needs to be grouped for similar details of individual
responses.

Data Preprocessing is a technique that is used to convert the raw data into a clean data set. In
other words, whenever the data is gathered from different sources it is collected in raw format
which is not feasible for the analysis.

Therefore, certain steps are executed to convert the data into a small clean data set. This
technique is performed before the execution of Iterative Analysis. The set of steps is known
as Data Preprocessing. It includes -

• Data Cleaning

• Data Integration

• Data Transformation

• Data Reduction

Dept. of ISE,Vemana IT Page 12 2019-2020

Bird Species Classification using Deep Learning Literature Survey

Data Preprocessing is necessary because of the presence of unformatted real-world data.

Mostly real-world data is composed of -

• Inaccurate data (missing data) - There are many reasons for missing data such as
data is not continuously collected, a mistake in data entry, technical problems with
biometrics and much more.

• The presence of noisy data (erroneous data and outliers) - The reasons for the
existence of noisy data could be a technological problem of gadget that gathers data, a
human mistake during data entry and much more.

• Inconsistent data - The presence of inconsistencies are due to the reasons such that
existence of duplication within data, human data entry, containing mistakes in codes
or names, i.e., violation of data constraints and much more.

Implementing:
In this work, a business intelligent model has been developed, to classify different Birds,
based on a specific business structure deal with Birds classification using a suitable Deep
learning technique. The model was evaluated by a scientific approach to measure accuracy.
We are using Convolutional Neural Network (CNN) to build our model.

Convolutional Neural Network

A convolutional neural network (CNN) is a special architecture of artificial neural networks,

proposed by Yann LeCun in 1988. CNN uses some features of the visual cortex. One of the
most popular uses of this architecture is image classification. For example Facebook uses
CNN for automatic tagging algorithms, Amazon — for generating product recommendations
and Google for search through among users’ photos.

Let us consider the use of CNN for image classification in more detail. The main task of
image classification is acceptance of the input image and the following definition of its class.
This is a skill that people learn from their birth and are able to easily determine that the image
in the picture is an elephant. But the computer sees the pictures quite differently:

Dept. of ISE,Vemana IT Page 13 2019-2020

Bird Species Classification using Deep Learning Literature Survey

Figure 5.0.2: Implementation

Instead of the image, the computer sees an array of pixels. For example, if image size is 300 x
300. In this case, the size of the array will be 300x300x3. Where 300 is width, next 300 is
height and 3 is RGB channel values. The computer is assigned a value from 0 to 255 to each
of these numbers. This value describes the intensity of the pixel at each point.

To solve this problem the computer looks for the characteristics of the base level. In human
understanding such characteristics are for example the trunk or large ears. For the computer,
these characteristics are boundaries or curvatures. And then through the groups of
convolutional layers the computer constructs more abstract concepts.

In more detail: the image is passed through a series of convolutional, nonlinear, pooling
layers and fully connected layers, and then generates the output.

The Convolution layer is always the first. The image (matrix with pixel values) is entered
into it. Imagine that the reading of the input matrix begins at the top left of image. Next the
software selects a smaller matrix there, which is called a filter (or neuron, or core). Then the
filter produces convolution, i.e. moves along the input image. The filter’s task is to multiply
its values by the original pixel values. All these multiplications are summed up. One number
is obtained in the end. Since the filter has read the image only in the upper left corner, it
moves further and further right by 1 unit performing a similar operation. After passing the
filter across all positions, a matrix is obtained, but smaller then an input matrix.

Dept. of ISE,Vemana IT Page 14 2019-2020

Bird Species Classification using Deep Learning Literature Survey

Figure5.0.3: Input neuron and first hidden layer

This operation, from a human perspective, is analogous to identifying boundaries and

simple colors on the image. But in order to recognize the properties of a higher level such as
the trunk or large ears the whole network is needed.

The network will consist of several convolutional networks mixed with nonlinear and
pooling layers. When the image passes through one convolution layer, the output of the first
layer becomes the input for the second layer. And this happens with every further
convolutional layer.

The nonlinear layer is added after each convolution operation. It has an activation function,
which brings nonlinear property. Without this property a network would not be sufficiently
intense and will not be able to model the response variable (as a class label).

The pooling layer follows the nonlinear layer. It works with width and height of the image
and performs a down sampling operation on them. As a result the image volume is reduced.
This means that if some features (as for example boundaries) have already been identified in
the previous convolution operation, than a detailed image is no longer needed for further
processing, and it is compressed to less detailed pictures.

Dept. of ISE,Vemana IT Page 15 2019-2020

Bird Species Classification using Deep Learning Literature Survey

Figure5.0.4: The Pooling layer and convolution layer

After completion of series of convolutional, nonlinear and pooling layers, it is necessary to

attach a fully connected layer. This layer takes the output information from convolutional
networks. Attaching a fully connected layer to the end of the network results in an N
dimensional vector, where N is the amount of classes from which the model selects the
desired class.

Analysis:

In this final phase, we will test our classification model on our prepared image dataset and
also measure the performance on our dataset. To evaluate the performance of our created
classification and make it comparable to current approaches, we use accuracy to measure the
effectiveness of classifiers.

After model building, knowing the power of model prediction on a new instance, is very
important issue. Once a predictive model is developed using the historical data, one would be
curious as to how the model will perform on the data that it has not seen during the model
building process. One might even try multiple model types for the same prediction problem,
and then, would like to know which model is the one to use for the real-world decision
making situation, simply by comparing them on their prediction performance (e.g., accuracy).
To measure the performance of a predictor, there are commonly used performance metrics,
such as accuracy, recall etc. First, the most commonly used performance metrics will be
described, and then some famous estimation methodologies are explained and compared to
each other. "Performance Metrics for Predictive Modelling In classification problems, the
primary source of performance measurements is a coincidence matrix (classification matrix

Dept. of ISE,Vemana IT Page 16 2019-2020

Bird Species Classification using Deep Learning Literature Survey

or a contingency table)”. Above figure shows a coincidence matrix for a two-class

classification problem. The equations of the most commonly used metrics that can be
calculated from the coincidence matrix are also given in Fig 2.7.

Figure 5.0.5: confusion matrix and formulae

As being seen in above figure, the numbers along the diagonal from upper-left to lower-right
represent the correct decisions made, and the numbers outside this diagonal represent the
errors. "The true positive rate (also called hit rate or recall) of a classifier is estimated by
dividing the correctly classified positives (the true positive count) by the total positive count.
The false positive rate (also called a false alarm rate) of the classifier is estimated by dividing
the incorrectly classified negatives (the false negative count) by the total negatives. The
overall accuracy of a classifier is estimated by dividing the total correctly classified positives
and negatives by the total number of samples.

5.1 System Architecture

System Architecture is an organized description that defines the structure, behavior and the
system views. The system architecture describes the major components, their relationships,
structures and how they interact with each other. Software architecture and design includes
several contributory factors such as Business strategy, quality attributes and many more.
Software Architecture and Design can be classified into two phases - Software
Architecture and Software Design. Software architecture refers to the fundamental
structure of a software system where each structure comprises of elements of the software,
the relationship between them and the properties of elements and the relations.
In architecture, the nonfunctional decisions are separated by the functional requirements.
Software design is the process by which a specification of the software artifact is created

Dept. of ISE,Vemana IT Page 17 2019-2020

Bird Species Classification using Deep Learning Literature Survey

with the intension to accomplish specific goals, using a set of primitive components and
subject to a set of constraints. In Design, functional requirements are accomplished.

The system architecture of the Recommender system, as shown in Fig 5.1, consists of
the following components – Admin, User, Database, Dataset. The Database consists of all
the required data for suggesting a proper restaurant to the user.Reviews,ratings are initially
collected from the websites and the official website to make sure that proper analysis can
be made.Algorithms required for analyzing the data in the database and providing points
based on the specified criteria are stored and are programmed to start when the user
requires the service. Admin has to update the details of the available restaurants to the
database for proper storage and sorting of data and admin should also be checking for
newly opening or opening soon restaurants to provide user with many options. To verify
the data provided by the official websites of the restaurants the admin has to survey the
places so that the fake data is not considered while allotting points. User can change his
preferances based on the ocations or place, so the system should take those preferances
before going through the data to provide suggestions. When the preferances of users match
with others the suggestions are provided based on the other users feedback.

Figure 5.1: System architecture

Dept. of ISE,Vemana IT Page 18 2019-2020

Bird Species Classification using Deep Learning Literature Survey

5.2 Data Flow Diagram

A data flow diagram (DFD) is a graphical representation of the "flow" of data through an
information system, modelling its process aspects. A DFD is often used as a preliminary step
to create an overview of the system without going into great detail, which can later be
elaborated.

DATAFLOW DIAGRAM LEVEL O

Figure 5.2: Dataflow diagram level 0

DATAFLOW DIAGRAM LEVEL 1

Figure 5.2.1: Dataflow diagram level 1

Dept. of ISE,Vemana IT Page 19 2019-2020

Bird Species Classification using Deep Learning Literature Survey

DATAFLOW DIAGRAM LEVEL 2

Figure 5.2.2: Dataflow diagram level 2

FLOWCHART: A flowchart is one of the seven basic quality tools used in project
management and it displays the actions that are necessary to meet the goals of a particular
task in the most practical sequence. Also called as process maps, this type of tool displays a
series of steps with branching possibilities that depict one or more inputs and transforms them
to outputs.

Dept. of ISE,Vemana IT Page 20 2019-2020

Bird Species Classification using Deep Learning Literature Survey

Figure 5.2.3: Flowchart

Dept. of ISE,Vemana IT Page 21 2019-2020

Bird Species Classification using Deep Learning Literature Survey

5.3 Use Case Diagram

Use Case Diagrams are used to capture the dynamic aspects of the system.
However, this definition is too generic to describe the purpose, as other four diagrams namely
activity, sequence, collaboration, and State chart also have the same purpose. The distinct
purpose of the use case diagrams is to gather the system requirements and the actors, thereby
specifying the events of the system and their flow.

Use Case diagrams are used to gather the system requirements which includes both the
internal and external influences. These requirements are mostly design requirements. Hence,
when a system is analyzed, use cases are prepared and actors are identified. When the initial
task is complete, use case diagrams are modelled to present the outer view.

In brief, the objectives of Use Case Diagrams can be stated as follows −

 It is used to gather the system requirements.

 It is used to get an external view of a system.

 It is easy to identify the external and internal factors which are influencing the system.

 It is easy to show the interactions between the actors and the system requirements.

Figure 5.3: Use case diagram

Dept. of ISE,Vemana IT Page 22 2019-2020

Bird Species Classification using Deep Learning Literature Survey

5.4 Modular Description

• Image Acquisition and Preprocessing

• Data Preparation and Model construction
• Model training

• Model testing and evaluation

Figure 5.4: CNN Architecture for Detecting Bird Images

• Image Acquisition and Preprocessing

In this module we will get the data from the online source. Further we will
resize the image for future use. Image resizing, or image scaling, is a geometric image
transformation which modifies the image size based on an image interpolation
algorithm. This image scaling process can increase or decrease the resolution of a
target image so that the absolute size of image data is adjusted.

Computers are able to perform computations on numbers and are unable to

interpret images in the way that we do. We have to somehow convert the images to
numbers for the computer to understand. The image will be converted to grayscale
(range of gray shades from white to black) the computer will assign each pixel a value
based on how dark it is. All the numbers are put into an array and the computer does
computations on that array. We then feed the resulting array for next step.

• Data Preparation and Model construction

Many a times, people first split their dataset into 2 — Train and Test. After this,
they keep aside the Test set, and randomly choose X% of their Train dataset to be the
actual Train set and the remaining (100-X)% to be the Validation set, where X is a

Dept. of ISE,Vemana IT Page 23 2019-2020

Bird Species Classification using Deep Learning Literature Survey

fixed number(say 80%), the model is then iteratively trained and validated on these
different sets. So we will follow the same method to prepare data for training and
testing phase.

We are building our model by using Convolutional neural network.

Convolutional neural networks (CNN) are a special architecture of artificial neural
networks, proposed by Yann LeCun in 1988. CNN uses some features of the visual
cortex. Now that we’re done pre-processing, we can start implementing our neural
network. We’re going to have 3 convolution layers with 2 x 2 max-pooling.

Max-pooling: A technique used to reduce the dimensions of an image by

taking the maximum pixel value of a grid. This also helps reduce over fitting and
makes the model more generic. After that, we add 2 fully connected layers. Since the
input of fully connected layers should be two dimensional, and the output of
convolution layer is four dimensional, we need a flattening layer between them. At the
very end of the fully connected layers is a softmax layer.

• Model training

After model construction it is time for model training. We were able to build
an artificial convolutional neural network that can recognize images. Split the dataset
into train and test dataset. Finally, we will build and train the model using training
dataset.

• Model testing and evaluation

Once the model has been trained it is possible to carry out model testing. During this
phase a test set of data is loaded. This data set has never been seen by the model and
therefore its true accuracy will be verified. Finally, the saved model can be used in the real
world. The name of this phase is model evaluation. This means that the model can be
used to evaluate new data.

Dept. of ISE,Vemana IT Page 24 2019-2020

Bird Species Classification using Deep Learning Literature Survey

CHAPTER 6

SUMMARY
We give the image of bird to our model it will take the image and do the pre-processing and
resize and then it will feed throw the image into conv2d layer and max pooling layer then
dropout layer finally it moves to flatten layer then output will come witch species bird this
photo belongs.

Future Scope:
Feature extraction is vital to the classification of relevant information and the differentiation
of bird species. We combined bird data from the Internet of Birds (IoB) and an Internet bird
dataset to learn the bird species.

Dept. of ISE,Vemana IT Page 25 2019-2020

Bird Species Classification using Deep Learning Literature Survey

CONCLUSION
This system comes under deep learning which is advanced technique at present. CNN is more
suitable for image processing especially in image classification. We conclude the
experimental result what we are getting from developed system is more accurate. The existing
system of Bird Species Identification many people try many different algorithms like Pose
Norm, Part-based R-CNN, Multiple granularity CNN, Diversied visual attention network
(DVAN) The deep LAC localization, alignment, and classification.Cannot properly represent
the diversity of Birds classes with complex intra-class variability and inter-class similarity.
Less effective for Birds classification compare to human.A novel method is proposed for
Birds face classification based on one of the popular convolutional neural network (CNN)
features. We are using CNN which can automatically extract features, learn and classify them.
CNNs use relatively little pre-processing compared to other image classification algorithms.
This means that the network learns the filters that in traditional algorithms were hand-
engineered. This independence from prior knowledge and human effort in feature design is a
major advantage in image classification.

Dept. of ISE,Vemana IT Page 26 2019-2020

Bird Species Classification using Deep Learning Literature Survey

REFERENCES
1. Likeability of Garden Birds: Importance of Species Knowledge & Richness in
Connecting People to Nature.
2. Image Net Large Scale Visual Recognition Challenge
3. Object-Part Attention Model for Fine-Grained Image Classification
4. Landmark Classification with Hierarchical Multi Modal Exemplar Feature
5. Object Detection with Deep Learning: A Review

Dept. of ISE,Vemana IT Page 27 2019-2020