PLANT DISEASE DETECTION USING

MACHINE LEARNING

Project report submitted in partial fulfillment of the

requirement for the degree of Bachelor of Technology

in

Computer Science and Engineering/Information

Technology

By

Vaibhav Sharma (191545)

Under the supervision of

Dr. Ruchi Verma
to

Department of Computer Science & Engineering and

Information Technology

Jaypee University of Information Technology

Waknaghat, Solan- 173234, Himachal Pradesh
 CERTIFICATE

I hereby declare that the work presented in this report entitled “ Plant Disease
Detection using Machine Learning” in partial fulfillment of the requirements
for the award of the degree of Bachelor of Technology in Computer Science
and Engineering/Information Technology submitted in the department of

Computer Science & Engineering and Information Technology, Jaypee

University of Information Technology, Waknaghat is an authentic record of my
own work carried out over a period from July 2022 to May 2023 under the
supervision of Dr. Ruchi Verma, Associate Professor(SG), Dept. CSE&IT.
I also authenticate that I have carried out the above mentioned project work
under the proficiency stream Machine Learning.
The matter embodied in the report has not been submitted for the award of any
other degree or diploma.

(Student Signature)
Vaibhav Sharma(191545)

This is to certify that the above statement made by the candidate is true to the
best of my knowledge.

(Supervisor Signature)
Dr. Ruchi Verma
Associate Professor (SG)
Department of CSE & IT

(i)
PLAGIARISM REPORT

(ii)
 ACKNOWLEDGEMENT

At the onset, I express my heartfelt thanks and gratefulness to God for his pure
and divine blessing that makes it possible for us to complete the project work
successfully within the right time. I am really humbled to do this endeavor
project under my respected professor and I wish my profound indebtedness to
Supervisor Dr. Ruchi Verma, Assistant Professor, Department of CSE & IT,
Jaypee University of Information Technology (JUIT), Waknaghat. She has
deep Knowledge of this project related stuff & her keen interest in guiding us
in the field of "Machine Learning" to carry out this project. Her endless
patience, scholarly guidance, perennial encouragement, constant and energetic
supervision, and valuable advice pertaining to many pre-published drafts have
made it possible for me to complete this project.

I would like to express my deep gratitude from the bottom of my heart to Dr.
Ruchi Verma, Department of CSE, for her generous help to finish my project.

I would also acknowledge each one of those individuals who have helped me
directly or indirectly in making this project a possibility. In this juncture, I
would like to thank the staff fraternity, individuals, both educating and
non-instructing, which have developed their convenient help and facilitated
my undertaking.

Vaibhav Sharma(191545)

(iii)
 TABLE OF CONTENT

Content Page No.

CERTIFICATE i
PLAGIARISM CERTIFICATE ii
ACKNOWLEDGEMENT iii
LIST OF FIGURES v
ABSTRACT vi
1. Chapter -1: INTRODUCTION 1-11
1.1 Introduction 1-3
1.2 Problem Statement 3-4
1.3 Objectives 4-6
1.4 Methodology 6-10
1.5 Organization 11
2. Chapter – 2: LITERATURE SURVEY 12-13
3. Chapter – 3: SYSTEM DEVELOPMENT 14-29
3.1 Analysis/Design/Development/Algorithm 14-15
3.2 Hardware requirements 15
3.3 Input requirements 15-18
3.4 Output requirements 18-21
3.5 Implementation 21-31
4. Chapter – 4: PERFORMANCE ANALYSIS 32-43
4.1 Formulae 32-33
4.2 Result Analysis 34-43
5. Chapter – 5: CONCLUSION 44-46
5.1 Conclusion 44-45
5.2 Future Work 46
6. Reference 47
7. Appendice 48

(iv)
 LIST OF FIGURES

Fig.no Figure name Page no.

Fig. 1.1 Infected Leaf 9

Fig. 1.2 Infected Leaf 9

Fig. 1.3 Infected Leaf 9

Fig. 3.1 Description of CNN 18

Fig. 3.2 Example of Clustering 22

Fig 3.3 Classification vs Regression 23

Fig. 3.4 Decision Tree for Playing Tennis 27

Fig. 3.5 Neural Network 29

Fig. 4.1 VGG-16 Epoch Results 34

Fig. 4.2 The results of VGG-16 with an accuracy of 94% 35

Fig. 4.3 The Confusion Matrix for VGG-16 Model 35

Fig. 4.4 The epochs for DenseNet Model 36

Fig4.5 The results for the DenseNet Model 37

Fig 4.6 Architecture of the InceptionV3 Model 37

Fig 4.7 The epochs running for InceptionV3 Model 38

Fig 4.8 Confusion Matrix for InceptionV3 Model 38

Fig 4.9 The results for InceptionV3 Model 39

Fig 4.10 The Results of Ensemble Learning on both the classes 41

Fig 4.11 Confusion Matrix for the Ensemble Learning Model 42

(v)
 ABSTRACT

Plant disease detection is a cutting-edge and enlightening system that helps

users learn about diseases, training, and other fascinating events happening in
their local area. This organization helps the local population stay informed
about activities in and around their town, region, or locale. This approach
requires both machine learning and image processing in order to function. The
accuracy of the results has been improved by using contemporary methods like
machine learning and deep learning algorithms. As a whole, random forests are
a learning technique for problems like classification, regression, and others that
work by building a forest of decision trees during the training period.
A component descriptor used in computer vision and image processing for
object detection is the histogram of oriented gradients (HOG). In this case, we
are using three component descriptors:
1. Hu moments
2. Haralick texture
3. Colour Histogram

Keywords: HOG, Object recognition, Random Forests, Image processing,

Histogram graph

(vi)
 Chapter-1 INTRODUCTION

1.1 Introduction
Plant disease detection is a cutting-edge and enlightening system that helps
users learn about diseases, trainings, and other fascinating events happening
nearby. This organisation helps the local population stay informed about
activities in and around their town, region, or locale. This approach requires
both machine learning and image processing in order to function. The user is
only allowed to add diseases related to his town, albeit they can be added for
any town. When a user adds an unsuitable, fictitious, or misused ailment, the
administrator will display it and take appropriate action. Only 500 words are
available for the user to lecture on diseases. The system gives a swipe to
advance to the next or earlier disease with transition features, and the
appearance and texture of the disease evaluation are exciting and
outstanding.Farmers in rural areas might believe it is difficult to distinguish the
diseases that might affect their harvests. They cannot easily visit the
agricultural office to find out what the infection might be. Our main goal is to
identify the disease that is introduced in a plant by observing its shape using
image processing and machine learning.
Pests and diseases harm crops or plant parts, reducing food production
and escalating food poverty. Toxic infections, inadequate disease management,
and significant climatic change are some of the major contributors to
decreasing food production. Numerous innovative technologies have emerged
in order to decrease post-harvest processing, improve agricultural
sustainability, and boost production. Numerous laboratory-based techniques,
such as polymerase chain reaction, gas chromatography, mass spectrometry,
thermography, and hyperspectral approaches, have been employed to identify
illnesses. These techniques are time-consuming and inefficient from a financial
standpoint. Recently, server-based and mobile-based methods have been used
to identify diseases. The high resolution camera, among other

1
things, makes it possible to automatically identify diseases.Using modern
techniques like machine learning and deep learning algorithms has increased
the accuracy of the outcomes. Numerous research have been carried out
utilising traditional machine learning techniques, including random forests,
artificial neural networks, support vector machines (SVM), fuzzy logic,
K-means method, and convolution neural networks, among others, for the
detection and diagnosis of plant illnesses. In general, random forests are a
learning method that develops a forest of decision trees during the training
phase and applies it to problems like classification, regression, and others.

A component descriptor used in computer vision and image processing for

object detection is the histogram of oriented gradients (HOG). In this case, we
are using three component descriptors:

1. Hu moments
2. Haralick texture
3. Color Histogram Hu moments is basically used to extract the shape of the
leaves.

To obtain the colour and texture of the leaves, haralick texture is utilised. A
histogram is a graph that shows how the colours in an image are distributed.
Farmers in rural areas might believe it is difficult to distinguish the diseases
that might affect their harvests. They cannot easily visit the agricultural office
to find out what the infection might be. Our main goal is to identify the disease
that is introduced in a plant by observing its morphology, processing its
images, and using machine learning. Pests and diseases cause crops or parts of
plants to be destroyed, which lowers food output and increases food insecurity.
Various less developed nations also have a lower level of expertise on the
prevention and control of illnesses and pests.
In the recent past, disease identification has been done using server-based and
mobile-based approaches. Automatic disease recognition is made possible by a

2
number of elements, including the high resolution camera, high performance
processing, and numerous built-in accessories. A portable system for detecting
plant diseases operates in an automated environment built with the Java
language. The app will operate quickly and smoothly. Because Google
Material Design is used, apps look more elegant and beautiful and provide a
positive user experience. It manages news, different categories, notifications,
and many other things whenever we want when admin and user are combined.
Good design and good programming are priorities. Using this app
allows us to save time and money while also allowing us to construct our own
unique or distinct types of apps based on our needs. Plant disease detection is a
cutting-edge and enlightening system that helps users learn about diseases,
trainings, and other fascinating events happening nearby. The local community
is assisted by this organization in staying informed about events taking place in
and around their town, area, or location. In order for this method to work, both
machine learning and image processing are necessary. The user is only allowed
to view diseases related to his town, though they can be added for any town.
When a user adds an unsuitable, fictitious, or misused ailment, the
administrator will display it and take appropriate action. Android Studio is the
front end, and SQL Server is the back end. To use this app, the user must
register with the system and may also update his information.

1.2 Problem Statement

To create a trustworthy and accurate system that can automatically detect and
diagnose plant illnesses in a quick and timely manner is the problem
statement for plant disease detection using machine learning. This system
should be able
to generalise to new locations and plant species while processing massive
amounts of data rapidly and accurately.

3
 The present methods for identifying plant diseases, such visual inspection or
laboratory analysis, take time, are subjective, and frequently need specialised
knowledge. Furthermore, these techniques might not be able to identify
diseases in the earliest stages, when treatments are most efficient.
The potential for machine learning-based methods to get over these
restrictions and offer an automated, objective solution for plant disease
diagnosis. However, the creation of such systems necessitates the
accessibility of sizable and varied datasets, the choice of suitable machine
learning algorithms, and the hyperparameter optimisation for the particular
issue at hand.

1.3 Objectives
1) Early Detection: Preventing the spread of plant diseases requires early
detection. By assisting in the early detection of plant illnesses, machine
learning algorithms enable farmers to respond quickly to limit additional
harm.

2) Accuracy: Based on the examination of sizable datasets of plant photos,

machine learning algorithms can precisely diagnose and categorize many
forms of plant diseases. This precision can decrease the possibility of a wrong
diagnosis and improve the management of plant diseases.

3) Efficiency: By automating the process of finding plant diseases, machine

learning algorithms may assist increase efficiency by saving time and effort.
This strategy can help farmers and experts in plant diseases swiftly monitor
broad regions of crops and spot possible outbreaks.

4) Cost-effectiveness: Developing cost-effective detection methods is important

for widespread adoption, especially in resource-limited agricultural systems.
The objective is to create affordable and accessible technologies that can be
utilized by farmers and agricultural practitioners.

4
1.4 Methodology

The following models have been used and trained for the detection of Plant Diseases
and the datasets have been taken from Kaggle.com, then the models with the highest
accuracy are selected and they undergo an ensemble model.

1) VGG-16 Model- The Visual Geometry Group (VGG) at the University of

Oxford created the convolutional neural network (CNN) model known as the
VGG16 in 2014. The moniker VGG16 refers to the deep learning
architecture's 16 layers.Multiple convolutional layers with tiny 3x3 filters
make up the VGG16 model.
2) DenseNet201-Researchers at Facebook AI Research presented the
DenseNet201 model, a deep convolutional neural network (CNN)
architecture, in 2017. It is a variation on the DenseNet (Dense Convolutional
Network) model, which was created to deal with the issue of vanishing
gradients in extremely deep neural networks.The DenseNet201 model's
architecture is made up of a number of dense blocks, which are collections of
convolutional layers with a same number of output channels.
3) InceptionV3- Researchers at Google introduced the InceptionV3 model, a
deep convolutional neural network (CNN) architecture, in 2015. It is an
adaptation of the Inception model, which was created to solve the issue of
computing effectiveness in extremely deep neural networks. The 48 layers of
the InceptionV3 model are organised into many Inception modules, which are
collections of convolutional layers with various filter sizes.

Ensemble learning is a machine learning technique that involves combining multiple

models or learners to make predictions or solve complex problems. The idea behind
ensemble learning is that by combining the predictions of multiple models, the
overall performance and accuracy can be improved compared to using a single model.
The individual models in an ensemble are often referred to as "base models" or "weak
learners." These base models can be of any type, such as decision

5
trees, neural networks, support vector machines, or any other machine learning
algorithm. Each base model in the ensemble is trained
independently on a subset of the training data or using a different algorithm variation
to introduce diversity.

There are several common techniques for combining the predictions of the base
models in an ensemble. The two most popular methods are:

1) Voting: In voting-based ensembles, each base model independently predicts

the target variable, and the final prediction is made by aggregating the
individual predictions. This can be done through majority voting (where the
most frequent prediction is selected) or weighted voting (where each model's
prediction is given a specific weight).

2) Averaging: Averaging-based ensembles compute the average or weighted

average of the predictions made by the base models. This approach is
commonly used when dealing with regression problems, where the final
prediction is a continuous value.

Ensemble learning offers several advantages over using a single model. It helps in
reducing overfitting by introducing model diversity and capturing different aspects of
the data. Ensemble models tend to have better generalization capabilities and can
handle complex patterns in the data. They are also more robust to noise and outliers
in the training data.

Some popular ensemble learning algorithms include Random Forest, AdaBoost,

Gradient Boosting, and Bagging. Each algorithm employs a specific strategy to train
and combine the base models effectively.

6
Overall, ensemble learning is a powerful technique that can significantly improve the
performance and accuracy of machine learning models by leveraging the collective
knowledge of multiple models.

Fig. Ensemble Learning through a flowchart

Ensemble methods can help reduce the bias introduced by individual models. If a
base model is biased in some way, combining it with other models that have different
biases can lead to a more balanced and unbiased overall prediction.

The dataset used is the Banana Leaf Dataset from Kaggle.com and it has 2000 images
across 2 classes of healthy and unhealthy leaves with 256x256 pixels of each leaf.
\

7
1.4.1 Data Augmentation
In machine learning, data augmentation is utilized to expand the size and
variety of a training dataset by generating synthetic data from existing data.
This method is particularly advantageous for detecting plant diseases, as there
may be a restricted number of images accessible for training.

There are several ways to perform data augmentation in the detection of plant
diseases using machine learning, including:

1. Image rotation: By rotating the image at various angles, new images

can be created from the existing data.
2. Image flipping: Flipping the image horizontally or vertically can create
new images.
3. Image scaling: Scaling the image up or down can create new images.
4. Image cropping: Cropping the image at different sizes and positions
can create new images.
5. Image translation: Shifting the image in different directions can create
new images.
6. Color augmentation: Changing the brightness, contrast, and saturation
of the image can create new images.

By applying these techniques, a larger and more diverse training dataset can be
created, which can improve the accuracy of the machine learning model in
detecting plant diseases. It's worth noting, however, that data augmentation
alone cannot guarantee good performance; other factors such as the choice of
algorithm, hyperparameters, and the quality of the original dataset also play
important roles.

8
Fig 1.1 Infected leaf

Fig 1.2 Infected leaf

Fig 1.3 Infected leaf

9
1.4.2 Data Acquisition

Acquiring data is an essential aspect of developing machine learning models

for detecting plant diseases. There are several methods available for data
acquisition, including capturing images of diseased plants using a
high-resolution camera in a greenhouse, field, or laboratory setting. Multiple
images should be taken from different angles and lighting conditions to create
a diverse dataset. Another option is to use existing datasets, such as
PlantVillage, Plant Pathology 2020, and FungiNet, which already contain
images of plants with and without diseases. Crowdsourcing platforms like
Zooniverse or Amazon Mechanical Turk can also be utilized to collect labeled
images of plant diseases from numerous contributors. This approach may
require some supervision to ensure accurate labeling and can be
time-consuming.

Engaging with plant pathologists or agronomists who specialize in plant

diseases can offer access to crucial data and expertise. They can advise on the
selection of images to collect, which diseases to prioritize, and how to label
images accurately. After acquiring the data, it's vital to correctly label the
images with the appropriate disease class. This process is typically
accomplished manually by human experts or with the aid of crowdsourcing
platforms. Once the images have been labeled, they are utilized to train a
machine learning model to effectively classify new, unseen images of diseased
plants.

10
1.5 Organization

The organisation of the report is as follows:

1) Chapter 1- Introduction
2) Chapter 2- Literature Survey 3)
Chapter 3- System Development 4)
Chapter 4- Performance Analysis 5)
Chapter 5- Conclusions

11
 Chapter-2 LITERATURE SURVEY

Machine learning approaches offer several advantages over traditional

methods in detecting plant diseases, including the ability to process large
amounts of data quickly and accurately, detect subtle changes in plant health,
and provide standardized diagnoses without subjectivity. Despite significant
advancements in machine learning for plant disease detection, challenges
remain, such as the requirement for large and varied datasets, generalization of
models to new environments, and the interpretation of models for
decision-making. However, the application of machine learning has the
potential to create efficient and effective tools for detecting and managing
plant diseases.

1) S. S. Sannakki Et al. [1] proposed a “Classification of Pomegranate

Diseases Based on Back Propagation Neural Network” which mainly works
on the method of Segment the defected area and color and texture are used
as the features. They have used neural networks here.
2) P. R. Rothe Et al. [2] introduced a” Cotton Leaf Disease Identification using
Pattern Recognition Techniques” which Uses snake segmentation, here Hu’s
moments are used as distinctive attribute. Active contour model used to limit
the vitality inside the infection spot, BPNN classifier tackles the numerous
class problems. The average classification is found to be 85.52%.
3) Aakanksha Rastogi Et al. [3] “ Leaf Disease Detection and Grading using
Computer Vision Technology &Fuzzy Logic”. K-means clustering used to
segment the defected area; GLCM is used for the extraction of texture
features, Fuzzy logic is used for disease grading. They used artificial neural
network (ANN) as a classifier which mainly helps to check the severity of the
diseased leaf.

12
4) Godliver Owomugisha Et al. [4] Automated Vision-Based Diagnosis
of Banana Bacterial Wilt Disease and Black Sigatoka Disease “Color
histograms are extracted and transformed from RGB to HSV. Peak
components are used to create max tree, five shape attributes are used
and area under the curve analysis is used for classification.They used
nearest neighbors, Decision tree, random forest, extremely randomized
tree, Naïve bayes and SV classifier. In seven classifiers extremely,
randomized trees yield a very high score, provide real time information
provide flexibility to the application.
5) Chunjiang Zhao Et al. [5] SVM-based Multiple Classifier System
for Recognition of Wheat Leaf Diseases,” Color features are
represented in RGB to HIS, by using GLCM, seven invariant moment
are taken as shape parameter. They used SVM classifier which has
MCS, used for detecting disease in wheat plant offline.

13
 Chapter-3 SYSTEM DEVELOPMENT

3.1 Analysis/Design/Development/Algorithm

Augmentation Process: The augmentation process involves increasing the size

of the dataset and introducing slight distortions to images to prevent overfitting
during training. Various transformation techniques, including affine and
perspective transformations and image rotations, can be used. Affine
transformations are useful for translations and rotations, where parallel lines in
the original image remain parallel in the output image. To find the
transformation matrix, three points from the original image and their
corresponding locations in the output image are required. Straight lines are not
affected by the transformation. Simple image rotations are commonly applied
during the augmentation process, with rotations occurring on different axes by
various degrees. Functional requirements in software engineering refer to
defining the function of a software system or its component.
Functional requirements refer to the specific tasks that a software system or its
components must perform, including calculations, data processing, and
technical details. In this context, the system needs to perform several tasks,
including taking in simple data in the form of .excel or .csv extensions,
performing data cleaning or preprocessing to prepare the data for further
processing, visualizing the data through charts or images of leaves, and
ultimately providing a predicted output based on the processed data.

Having appropriate datasets is crucial for all stages of object recognition

research, including the training phase and evaluating the performance of
recognition algorithms. The dataset used in this research was collected from
the Internet and included images of plant diseases and healthy leaves, as well
as background images. The images were grouped into fifteen different classes,
including thirteen classes representing plant diseases and one class

14
representing healthy leaves. Additionally, a class for background images was
added to improve classification accuracy.
To distinguish the leaves from the surrounding, a deep neural network
was trained on the dataset, including the background images taken from the
Stanford background dataset. In systems engineering and requirements
engineering, non-functional requirements refer to the criteria that can be used
to judge the operation of a system, rather than specific behaviors, in contrast to
functional requirements that define specific behavior or functions. The plan for
implementing functional requirements is detailed in the system design, while
the plan for implementing non-functional requirements is detailed in the
system architecture. Other terms for non-functional requirements include
"constraints", "quality attributes", "quality goals", "quality of service
requirements", and "non-behavioral requirements".

3.2 HARDWARE REQUIREMENTS

Processor : Any Processor

RAM : more than 2GB
Hard Disk : 10 GB
Input device : Standard Keyboard and Mouse Output device : High Resolution
Monitor
Memory : 1GB
RAM Requirements :Different plant leaf images (healthy and affected) 3.2.2
SOFTWARE
REQUIREMENTS Operating system : Windows XP and above, Windows 7
IDE : Python IDLE
Data : Leaf Dataset Modules : Numpy, pandas, matplotlib, sklearn Processor
:Ip3 processor

15
3.3 INPUT REQUIREMENTS

Functional requirements in software engineering define the function of a software system or

its components. Such requirements specify a set of inputs, behavior, and outputs that may
include calculations, technical details, data manipulation, processing, and other
functionalities that explain what the system aims to accomplish. Use cases capture the
behavioral requirements that detail how the system uses the functional requirements. In this
case, the system is expected to perform the following tasks: (1) acquiring simple data in an
Excel or CSV format, (2) performing data cleaning or preprocessing to transform raw data for
another processing procedure, (3) generating visual representations such as charts or images
to display information, and (4) producing predicted output.

Techniques used:

Image processing: The user has to record into the system by giving his own information.

HOG: The registered user has to login into the app to add or see the disease, the user is
recorded once he logs in until he logs out holding his time to login every other time.

Machine learning: The user is given chance to add disease choosing a town or area to refer to
also add an image related to the news.

CNN: The user can see the disease based on the zone the user will select.

16
Artificial Intelligence : User can also view the list of disease added by him and
action taken by admin if any.

Several disease: If the user finds any disease is not real or unpleasant he can
say it to the admin, admin will take some act.

Random forest: Admin should login into the app to check disease or report
submitted by the user.

SMV: Admin can see all the disease of all zones.

Viewing diseases: Admin is answerable to check any disease is been conveyed

to him to cross checked.

To conduct object recognition research, appropriate datasets are essential

throughout the process of training and evaluating recognition algorithms. In
the case of a plant disease dataset, images were collected from various online
sources using multiple languages to search for specific plant names and
diseases. The images were categorized into fifteen classes, with thirteen
classes representing diseased plants and one class for healthy leaves. An
additional class for background images helped improve classification accuracy
by training a deep neural network to distinguish leaves from the surrounding
environment. These background images were obtained from the Stanford
background dataset. Once a simple dataset is obtained in .excel or .csv format,
data cleaning and preprocessing are performed to prepare the raw data for
further processing. Following this step, visualization is done to represent the
information as an image or chart. Finally, the predicted output is generated.

17
 Fig. 3.1 Description of CNN

3.4 OUTPUT REQUIREMENTS

In the field of systems engineering and requirements engineering, a type of
requirement known as output requirement defines criteria that can be used to
evaluate the operation of a system, as opposed to specifying specific
behaviors. This differs from functional requirements, which describe specific
behaviors or functions of a system. The process of implementing functional
requirements is described in the system design, while the implementation of
non-functional requirements is detailed in the system architecture.
Non-functional requirements may also be referred to as "constraints", "quality
attributes", "quality goals", "quality of service requirements", or
"non-behavioral requirements". Various quality attributes may be considered
as part of non-functional requirements.

● ACCESSIBILITY: Accessibility refers to the extent to which a

product, device, service, or environment can be easily used and
accessed by the maximum number of people. In the context of our

18
 project, it is accessible to anyone who has Python installed on their
system.

● MAINTAINABILITY: In the field of software engineering,

maintainability refers to the level of ease with which a software
product can be altered to address defects, meet new requirements, or
add new features/data. In the case of this specific project, new images
of leaves can be easily added as new features/data.

● SCALABILITY: The system has the ability to handle an increase in

total throughput when there is an increase in load, provided that
additional resources, typically hardware, are added. Moreover, the
system is capable of functioning properly even under conditions
involving large datasets and low bandwidth.

● PORTABILITY:Portability is an important characteristic of

high-level programming. It refers to the ability of software code to be
reused when moving a program from one environment to another,
rather than having to create new code. Our project is portable in that it
can be executed under different operating conditions, as long as the
minimum requirements are met and a suitable dataset is available.
Specifically, our application is written in the Python programming
language and packaged as an APK file for use on Android devices. The
APK contains compiled Java code as well as other resources like
scripts and images.

● ACTIVITIES:Artificial intelligence and machine learning are a

subset of computer science that deals with the development of
intelligent algorithms that can learn from data. These algorithms can be
integrated into software applications and typically occupy a section of

19
 the application's codebase. In the case of mobile devices, the AI/ML
functionality often takes up the entire screen and is responsible for
managing the user interface and interactions with the device's screen.

● FRAGMENTS: In Android, a Fragment is a modular section of a

user interface that can be combined with other fragments to create a
complete UI. A fragment can occupy only a part of the screen or the
entire screen, depending on the size and complexity of the UI.
Fragments are used to build flexible UI designs that can be reused
across different activities. Fragments can contain views and view
groups, and can be added or removed dynamically at runtime to create
rich and responsive user experiences.

● SMV AND HOG: In Android, GUI elements are called Views, such
as a TextView displaying text or a Button that users can click on. View
Groups are containers for Views. A ViewGroup can contain a
collection of Views together, and Views and View Groups can be
nested within each other. Fragments and Activities can use XML files
to define their layout and content. The layout XML files define the
GUI elements included in an Activity or Fragment, as well as the
layout of those GUI elements (size, margins, padding, etc.).

● INTENTS: System intents are minor substances that an activity can

permit to the python operating organization, to tell the operating
system that certain other act or action is required.

● WIDGETS: Machine widgets refer to graphical user interface

elements that can be displayed independently of an activity. In the case

20
 of plant disease detection, these widgets are used to display and
visualize diseases as part of a detection request. In some cases, views
of plant diseases may also be referred to as widgets.

● SERVICES: Even if no requests are displayed, family processes for

plant disease can be implemented on a machine device. A user barrier
is not what services want. A detection service might, for instance,
periodically hold up data or check a remote server for the presence of
illnesses.

3.5 LANGUAGE SPECIFICATION

1) PYTHON

Python is a popular high-level programming language known for its

readability and use of significant whitespace. It was created in 1991 by
Guido van Rossum and supports both small and large scale
programming. C Python is the reference implementation of Python and
is open source with a community-based development model. Python
supports many programming paradigms including object-oriented,
structured, functional, and aspect-oriented programming. It also uses
dynamic typing and reference counting with a garbage collector for
memory management. Python has built-in support for functional
programming features like filter(), map(), and reduce() functions, as
well as list comprehensions, dictionaries, and sets. The language has a
core philosophy summarized in "The Zen of Python" document which
includes aphorisms like "Simple is better than complex" and "Explicit
is better than implicit."

21
2) CLUSTERING
A cluster refers to a set of data objects that share similar
characteristics, while dissimilar data objects belong to different
clusters. Clustering refers to the process of organizing abstract objects
into similar groups. A cluster of data objects is treated as a single
group. During cluster analysis, data is divided into groups based on
their similarity and given labels. Unlike classification, clustering is
adaptive and can identify features that distinguish different groups. The
k-means algorithm will be used to cluster crime data in the proposed
system.

Fig 3.2 Example of Clustering

3) CLASSIFICATION

This method is employed to categorize various types of data. The way

this works is comparable to clustering. It divides data records into
several segments, also referred to as classes. In contrast to clustering,
we are aware of the many clusters here. For instance, Outlook email
uses an algorithm to classify emails as real or spam.

22
 Fig 3.3 Classification versus Regression

4) DECISION TREE
This method plays a crucial part in machine learning since it is simpler
for people to comprehend. The decision tree starts with a root that is a
straightforward query. The questions on the root node of the decision
tree may lead to more questions because they are open to multiple
replies. As a result, the decision tree's nodes continue to grow. We're
finally allowed to decide on it definitively.

Advantages:

1) More efficient.

2) Better crime monitoring systems.

3) Reduces the costs of storage, maintenance and personnel.

23
 4) It reduces the time complexity of the system.

5) System that has a simpler architecture to understand.

6) Processing of large amount of data becomes easier.

JDK

(Java SE Development Kit) Contains a full JRE (Java Runtime Environment)

in addition to tools for creating, troubleshooting, and keeping track of Java
applications.

To create and run Java programmes and applets, JDK is necessary. Five
categories are used to group JDK tools:
Simple Tools Tools for Remote Method Invocation (RMI) Internationalisation
Tools for Security IDL Java Tools

JDK Starter Tools The Java Development Kit is built on these tools.

javac

The .java file is compiled using the Java programming language's compiler,
javac. It generates a classfile that may be used with a Java command to run it.
Java The java command can be used to execute a Java programme after a class
file has been created.The command prompt is used to run both.

24
java

It is the extension for text files that include Java source code. After it has been
coded and saved, the javac compiler is used to produce.class files. The Java
command can be used to launch the Java programme as the.classfiles are
prepared.

javadoc

JavaDoc is an API documentation generator for the Java programming

language that creates documentation from Java source code in HTML format.

appletviewer

With the standalone command-line programme appletviewer, Java applets can

be run and debugged without the need for a web browser. jar The jar file
format (manage Java archive) is a collection of class, text, image, and sound
files for a Java programme or applet that is compressed into a single file.

3.6 ECLIPSE SDK FEATURES

The Eclipse Project offers a range of resources and can be downloaded from its
downloads page. Its sixteenth annual release provides several new features for
the Platform and Equinox, Java developers, and plug-in developers. Eclipse is
an IDE used for computer programming, particularly as a Java IDE, and
includes a customizable plug-in system within its base workspace.

25
Additionally, the K means clustering algorithm is a simple unsupervised
learning algorithm used for cluster analysis and is illustrated in a flowchart.

3.7 Decision Tree

This method plays a crucial part in machine learning since it is simpler for
people to comprehend. The decision tree starts with a root that is a
straightforward query. The questions on the root node of the decision tree may
lead to more questions because they are open to multiple replies. As a result,
the decision tree's nodes continue to grow. We're finally allowed to decide on
it definitively.

Advantages:

1)More efficient.

2)Better crime monitoring systems.

3)Reduces the costs of storage, maintenance and

personnel. 4)It reduces the time complexity of the system.

5)System that has a simpler architecture to understand.

6)Processing of a large amount of data becomes easier.

26
 Fig. 3.4 Decision Tree for Playing Tennis

3.8 Support Vector Machine (SVM)

They have remained the go-to approach for an algorithm with great
performance and little tuning since they were developed in the 1990s. Support
vector machines are a class of supervised learning models with corresponding
learning algorithms used in machine learning to analyse data used in
classification and regression analysis. Unsupervised learning is not possible
with unlabeled data. In a high-dimensional, infinite-dimensional space, it
creates a hyperlane and a group of hyperlanes that can be utilised for various
tasks, such as outlier detection.

27
Advantages

1) Effective in spaces with high dimensions.

2) Still useful in situations where there are more dimensions than

samples. 3) It also uses memory effectively.

4)Versatile.

5) Support vector machine (SVM) evaluates the data, classifies it, and then
does the regression.

6) The two primary characteristics for diagnosing the condition are accuracy
and detection time. SVMs, or support vector machines, boost recognition
rates.

3.9 NEURAL NETWORKS

The most common architecture, multilayer perceptron neural networks, and
Self-Organizing maps are the foundations of the illness detection systems.A
hardware or software system known as a neural network is modelled after how
neurons in the human brain function. Artificial neural networks are another
name for neural networks. Neural networks are a type of machine learning
model that are inspired by the structure and function of the human brain. They
are composed of layers of interconnected nodes or neurons, each of which
performs a simple mathematical operation on its inputs and passes the result
on to the next layer. By adjusting the strength of the connections between

28
neurons during training, neural networks can learn to recognize patterns in
data and make predictions or classifications.

Fig. 3.5 Neural network

3.10 FUNCTIONALITIES USED

The connectivity among different plants is established through a series of

activities. Initially, a picture of the plant leaf is captured in the first activity.
The second activity involves image processing to analyze the captured image.
A binary code of 1 and 0 is generated to indicate whether the leaf is healthy or
unhealthy. If the leaf is healthy, it will be marked as 1, and if not, it will be
marked as 0. The next step involves using machine learning algorithms to
determine the type of disease that has affected the plant if it is unhealthy. Upon
selecting the appropriate disease category, the final activity displays the
disease name, allowing the farmer to apply appropriate fertilizer to the plant to
eradicate the disease and pests. The last screen includes a message thanking
the user for reading and an option to go back and read about other diseases.

29
3.11 KEY ISSUES AND CHALLENGES

Detecting wheat leaf diseases is important for crop health and productivity, but
current machine learning methods face challenges. Digital images of wheat
leaves are captured to measure their shape, size, and texture through feature
extraction. Yellow rust disease caused by Puccinia striiformis Tritici produces
yellow spores on leaves during winter and spring, and new protection methods
like BTH aim to provide inherent disease resistance mechanisms. Seed-borne
diseases can reduce yield and quality of grains. Machine vision uses color and
geometric features for identification, but KNN classification is not suitable for
large applications due to distance calculations. Histogram equalization
enhances image contrast for better human perception and is used in wheat and
plant disease applications. Hyperspectral data is collected for healthy and
diseased wheat to build SVR models for disease index inversion and Random
Forest classification of wheat leaf diseases.

Although the use of machine learning for plant disease detection shows
promising results, there are still some key issues and challenges that need to be
addressed.

1) Lack of Standardized Datasets: One of the major challenges is the lack

of standardized datasets that can be used for training and testing
machine learning models. This can lead to bias and inaccurate results,
as different datasets may have different characteristics and variations.

2) Limited Availability of Expertise: Machine learning requires expertise

in both computer science and plant pathology. The availability of
individuals with expertise in both areas is limited, which can hinder the
development and implementation of accurate models.

30
 3) Variability in Environmental Conditions: Environmental factors such as
lighting, humidity, and temperature can have a significant impact on
plant health and appearance. Variability in these factors can make it
challenging to develop accurate machine learning models that can
detect diseases across different environments.

4) Limited Access to High-Quality Imaging Devices: High-quality

imaging devices such as hyperspectral cameras or drones can provide
detailed and accurate images of plants. However, these devices can be
expensive and may not be accessible to all growers or researchers,
limiting the data that can be collected for machine learning models.

5) Computational Complexity: Machine learning models can be

computationally complex, requiring significant processing power and
memory. This can make it challenging to develop models that can be
implemented in low-resource settings.

6) Lack of Integration with Decision-Making Tools: While machine

learning models can accurately detect plant diseases, they may not
provide actionable recommendations for growers or decision-makers.
There is a need for integration with decision-making tools to ensure
that the results from machine learning models can be effectively
translated into action.

Model Development

• Analytical
• Computational
• Experimental
• Mathematical

31
Chapter 4- PERFORMANCE ANALYSIS

The application of different models for prediction has provided the results
in different forms such as a confusion matrix, a comparison graph, and a
bar graph to compare the Accuracy, Precision, F1- Score and Recall of
different models. These values have been figured out using the formulae
given below.

4.1 Formulae

1) St = Tp / (Tp + Fn) Sensitivity (St) = True Positives / (True Positives +

False Negatives) In simple terms, Sensitivity quantifies the proportion
of actual positive instances that are correctly classified as positive by
the model, out of all positive instances.
2) Sf = Tn/ (Tn + Fp) Specificity (Sf) = True Negatives / (True Negatives
+ False Positives) In simple terms, Specificity quantifies the
proportion of actual negative instances that are correctly classified as
negative by the model, out of all negative instances.
3) TPR = Tp/ (Tp + Fn)

True Positive Rate is the number of instances that were predicted to be

True and correctly identified as True.

4) FPR = Fp/ (Fp + Tn)

False Positive Rate = False Positives / (False Positives + True

Negatives)

In simple terms, FPR quantifies the proportion of instances that are

truly negative but are incorrectly classified as positive by the model,
out of all actual negative instances.

32
5) A = (Tp + Tn)/ (Tp + Tn + Fp + Fn)

This is the formula for calculating the accuracy of the model. In

simple terms, accuracy quantifies the proportion of correctly classified
instances out of the total number of instances. It provides an overall
measure of how well the model performs in terms of correctly
predicting both positive and negative instances.

6) Precision(P) = Tp/ (Tp + Fp)

Precision = True Positives / (True Positives + False Positives) In

simple terms, precision measures how many of the instances that the
model predicted as positive are actually positive. It represents the
model's ability to avoid false positives and is indicative of its accuracy
when identifying positive instances.

7) Recall(R) = Tp/(Tp + Fn)

Recall = True Positives / (True Positives + False Negatives) In simple

terms, recall quantifies the model's ability to correctly identify positive
instances and captures the proportion of actual positives that were
correctly classified as positive. It is a measure of the model's
completeness or the proportion of relevant instances that were
successfully retrieved.

8) ERR = (Fp + Fn)/ (Tp + Tn + Fp + Fn)

ERR = (False Positives + False Negatives) / (True Positives + False

Positives + True Negatives + False Negatives) ERR (Error Rate) is a
performance metric used in machine learning and classification tasks
to measure the overall classification error of a model. It represents the
proportion of misclassified instances, both false positives and false
negatives, out of the total number of instances.

33
4.2 Result Analysis

The dataset is the Banana Leaves dataset, taken from Kaggle.com with
2000 different images, in two classes of Healthy and Unhealthy
leaves, and the following models were applied on the dataset, the
results are as follows.

1)VGG-16 Model

In Python, a file is regarded as a module. The module must be

imported using the import keyword before being used. By importing
the module, the function or variables existing in the file can be utilized
in another file. Instead of importing the entire module, we can merely
import the necessary functions and variable names from the module.
The "from" keyword can be used to import only certain items when
we only want those items imported. By using import and the module
name, i.e., the filename or the library to be used, we can import the
entire module. A Python library is a term that is widely used to
describe a group of linked modules. It contains code bundles that can
be used with many different software.

Fig 4.1 VGG-16 Epoch Results

Overfitting is a major issue when neural networks are trained using

sample data. A neural network model will learn very precise patterns
in the sample data if more training epochs are employed than are

34
necessary. As a result, the model won't fit the new data set properly.On
The training data set (sample data), this model achieves great
accuracy;however, on the test data set, it does not. In other words, by
overfitting the training set, the model loses its capacity to generalize.

Fig 4.2. The results of VGG-16 with an accuracy of 94%

The Accuracy achieved by the VGG-16 model is 94%. When the

result might potentially comprise two or more classes, it is a technique
to evaluate how effectively a deep learning categorization system
works. The table contains four possible combinations of predicted and
actual values. This program can be very effectively used to evaluate
AUC-ROC curves, recall, precision, specificity, accuracy, and—most
crucially—all of the aforementioned. The dataset was divided into two
classes the first being the healthy class to train the model and the
second being unhealthy leaves to check on the accuracy of the model.

Fig 4.3 The Confusion Matrix for VGG-16 Model

35
The confusion matrix representing the True Positive, True Negative,
False Positive and False Negative values across the two curves of
Predicted values and Actual values. The True Positive value in this
case is 5.0, meaning that of all the values in the dataset the predicted
value was true as well as the actual value. The True Negative value is
53.0 meaning that the model predicted the value to be negative and the
actual value was negative. False Positive value is 0.0 meaning that the
predicted no value was predicted to be false and came out as true. The
False Negative value is 4.0 meaning 4 of the values were predicted to
be true but the actual value was false.

2) DenseNet201

Fig 4.4 The epochs for DenseNet Model

The DenseNet Model stopped after 12 epochs due to early stopping.

If we let a complex model train long enough on a given data set it can
eventually learn the data exactly. Given data that isn’t represented in
the training set, the model will perform poorly when analyzing the
data (overfitting). Conversely if the model is only trained for a few
epochs, the model could generalize well but will not have a desirable
accuracy (underfitting). Hence the model stops early to prevent it from
learning the exact data.

36
 Fig 4.5 The results for the DenseNet Model with an accuracy of 92%

The accuracy of the DenseNet Model is 92% , with the Recall being
1.00, meaning that the False Negative value was 0, which means that
there was no value where the model predicted the value to be False
and the actual value was True.The Precision is 0.91 which means that
the False Positive was greater than 0 meaning that there was a False
value that was predicted to be True. The F1 Score is 0.95 which is the
harmonic representation of both Precision and Recall, being this high
means that the model is working well.

3) InceptionV3 Model

Fig 4.6 Architecture of the InceptionV3 Model

37
 Fig 4.7 The epochs running for InceptionV3 Model

The InceptionV3 Model ran all the epochs without stopping early
meaning that there was enough variation in the dataset for the model
to run all the way through because the model stops early if there is not
enough variation to prevent overfitting.

Fig 4.8 Confusion Matrix for InceptionV3 Model

38
The True Positive value for this model is 4.0 which means that there
were 4 cases that were predicted to be True and their actual value was
True, there are 53.0 cases of True Negative meaning they were
predicted to be False and the actual value was False.The False
Negative value is 1.0 meaning that one value was predicted to be false
but it turned out to be True.

Fig 4.9 The results for InceptionV3 with an accuracy of 90%

The accuracy for the InceptionV3 Model is 90%, with the Precision
being 91% and the Recall being 98% which means neither the False
Positive was 0 nor the False Negative, therefore there were a few
cases where the model predicted the wrong value. The F1- Score is
95% which is the harmonic representation of both the Precision and
Recall, meaning that the model is giving a majority correct prediction.

39
 Model Precision Recall F1-Score Accuracy

VGG-16 0.93 1.0 0.96 94

DenseNet201 0.91 0.95 0.95 92

InceptionV3 0.91 0.98 0.95 90

Table 1. Comparison of results between different models with the Banana Leaf Dataset

The highest accuracy is 94% achieved by the VGG-16 Model.

Graph 1. Graph showing results on the Banana Leaf Dataset

The above graph compares the Precision, F1- Score and the Recall of
all three models and the precision is the highest for the VGG-16
Model, the Recall is the highest for the VGG-16 Model, InceptionV3
Model has the highest F1-Score and overall DenseNet201 has the
highest accuracy.
 Applying Ensemble Learning

In ensemble learning, the averaging technique is a common method

used to combine the predictions of multiple base models. It is
primarily employed when dealing with regression problems, where the
goal is to predict a continuous value rather than a class label. The
averaging technique works by taking the average of the predictions
made by each base model in the ensemble. This can be done using a
simple arithmetic mean or by applying weighted averaging, where
each model's prediction is assigned a specific weight.

We have used the two models with the highest accuracy and applied
ensemble learning to it, the models being DenseNet201 and VGG-16
with the Banana Leaf Dataset.

Fig 4.10 The Results of Ensemble Learning on both the classes

Class Precision Recall F1-Score Accuracy

0 0.85 1.0 0.92

1 0.00 0.00 0.00 85

Table 2. The Results of Ensemble Learning on both the classes

41
 Macro average calculates the Precision, F1- Score and Recall for each
class and then finds the average by combining it with the other class
and calculating the average of both. This helps provide a definitive
view of both the classes, giving equal weight to each class. Weighted
average Precision is calculated by taking the sum of all True Positives
and then divided it by the sum of all True Positives and False Positives
of both the classes which gives a definite view of the whole dataset.
The same happens with Weighted average recall and F1-Score. The
accuracy is 85% for this combined model as it takes out the
redundancies in this previous models and gives a more precise
viewpoint of the situation.

Fig. 4.11 The Confusion Matrix Representing for the Ensemble Learning Model

The confusion matrix representing the True Positive, True Negative,

False Positive and False Negative values across the two curves of
Predicted values and Actual values. The True Positive value in this
case is 5.0, meaning that of all the values in the dataset the predicted
value was true as well as the actual

42
value. The True Negative value is 53.0 meaning that the model
predicted the value to be negative and the actual value was negative.
False Positive value is 0.0 meaning that the predicted no value was
predicted to be false and came out as true. The False Negative value is
4.0 meaning 4 of the values were predicted to be true but the actual
value was false.

43
 Chapter- 5 CONCLUSIONS

After applying ensemble learning, there are several potential

conclusions that can be drawn based on the results and performance of
the ensemble model. These conclusions may vary depending on the
specific problem, the ensemble technique used, and the evaluation
metrics employed. Here are the conclusions that can be made:

1) Improved performance: Ensemble learning leads to improved

performance compared to using a single model. If the
ensemble model outperforms the individual base models, it
suggests that the combination of diverse models has effectively
captured different patterns and achieved better predictive
accuracy or generalization. The use of Averaging Technique
has given a better view of the whole dataset and prevented
overfitting as the model had gotten used to the dataset but with
Averaging Technique this problem was prevented.
2) Reduced overfitting: Ensemble learning helped reduce
overfitting, particularly if the individual base models is prone
to overfitting the training data. If the ensemble model shows
improved performance on unseen data or during
cross-validation compared to individual models, it indicates
that the ensemble has successfully mitigated overfitting.
3) Robustness to noise and outliers: Ensemble models are more
robust to noise and outliers in the data compared to single
models. If the ensemble model exhibits better performance in
the presence of noisy or outlier data points, it suggests that the
ensemble has effectively learned to ignore or handle these
anomalies. The anomalies can cause the model to give
exaggerated performances and give very high accuracies which
was prevented by the use of this Ensemble Model.

44
4) Model diversity: Ensemble learning relies on the diversity of
the base models to improve performance. Analyzing the base
models' diversity, such as their differences in training
algorithms, hyperparameters, or feature subsets, can provide
insights into the effectiveness of ensemble learning. If the base
models are diverse and yet contribute positively to the
ensemble, it indicates the successful utilization of ensemble
techniques

In conclusion, using machine learning algorithms to identify

plant diseases has the potential to completely transform the
plant health management industry. Machine learning
algorithms can accurately diagnose and categorize plant
diseases using a variety of data sources, including pictures,
sensor data, and environmental characteristics. This technology
has several benefits, such as increased accuracy, early
detection, scalability, and the capacity to manage intricate data
interactions.

The potential for advancements in areas like the fusion of

various data sources, real-time monitoring and early warning
systems, transfer learning, deep learning for image analysis,
and integration with decision support systems make ML-based
disease detection look promising in the future. These
advancements may result in more specialized and proactive
illness management approaches, allowing for prompt
treatments and reducing financial losses.

Overall, the application of machine learning algorithms for

plant disease detection has the potential to transform
agriculture by enabling early and precise disease identification,
optimizing resource allocation, and encouraging sustainable
and effective farming methods. The creation of creative
methods for enhancing agricultural yields, lowering pesticide
facilitated by continued developments in this field.
5.1 Future Scope

For the future of this project, the best way to go forward with it
would be the use of more Machine Learning Models and
combining their results using Ensemble learning or the use of
Deep Learning Models which can bring a huge change to the
accuracy and prediction rate of this project. Moreover the use
of different datasets can also help, using the datasets for
different species of leaves plants. The use of drones in
photographing leaves can change the game for detection of
plant diseases as the use of drones combined with machine
learning algorithms has emerged as a powerful tool for several
advantages in this context, including the ability to capture
high-resolution aerial imagery, cover large areas efficiently,
and provide real-time data collection. When combined with
machine learning algorithms, these advantages can
significantly enhance disease detection and management
efforts.
Drones can cover large areas of agricultural land quickly and
efficiently, allowing for the collection of real-time data on
plant health. This rapid data acquisition enables early detection
and timely intervention, reducing the spread and impact of
diseases.

Machine learning algorithms, such as convolutional neural

networks (CNNs), can be trained to analyze drone-captured
images and identify disease symptoms. By training the
algorithms on large datasets of annotated images, they can
learn to recognize various diseases. interventions in disease
management. By identifying and mapping disease hotspots,
farmers can apply treatments only where necessary, reducing
the use of pesticides and optimizing resource allocation.

46
 REFERENCES

1) Ghaiwat Savita N, Arora Parul. Detection and

classification of plant leaf diseases using image
processing techniques: a review. Int J Recent Adv Eng
Technol 2014. (Online).
2) Dhaygude Sanjay B, Kumbhar Nitin P. Agricultural
plant leaf disease detection using image processing. Int
J Adv Res Electr Electron Instrum Eng 2013.
3) Mrunalini R Badnakhe, Deshmukh Prashant R. An
application of K-means clustering and artificial
intelligence in pattern recognition for crop diseases. Int
Conf Adv Inf Technol 2011 IPCSIT.
4) Arivazhagan S, Newlin Shebiah R, Ananthi S, Vishnu
Varthini S. Detection of unhealthy region of plant
leaves and classification of plant leaf diseases using
texture features. Agric Eng Int CIGR 2013.
5) Kulkarni Anand H, Ashwin Patil RK. Applying image
processing technique to detect plant diseases. Int J Mod
Eng Res 2012.
6) Bashir Sabah, Sharma Navdeep. Remote area plant
disease detection using image processing. IOSR J
Electron Commun Eng.
7) Naikwadi Smita, Amoda Niket. Advances in image
processing for detection of plant diseases. Int J Appl
Innov Eng Manage 2013.
8) Patil Sanjay B et al. Leaf disease severity measurement
using image processing. Int J Eng Technol 2011.
9) Chaudhary Piyush et al. Color transform based
approach for disease spot detection on plant leaf. Int
Comput Sci Telecommun 2012.

47
 APPENDICE

Here are some additional details that could be included in an appendix for the detection of plant
diseases using machine learning:
1) Types of machine learning algorithms: There are several types of machine learning algorithms
that can be used for plant disease detection, including supervised learning, unsupervised learning,
and reinforcement learning. Each algorithm has its own strengths and weaknesses and is suited to
different types of data and problems.

2) Performance metrics: There are several performance metrics that can be used to evaluate the
performance of a machine learning model for plant disease detection, including accuracy,
precision, recall, F1-score, and area under the curve (AUC). These metrics can help determine the
effectiveness of the model and identify areas for improvement.

3) Image preprocessing techniques: Image preprocessing techniques such as normalization, resizing,

and histogram equalization can be used to improve the quality and consistency of the images in
the dataset. These techniques can help reduce noise and variability in the images, making it easier
for the machine learning model to learn the patterns associated with diseased plants.

4) Transfer learning: Transfer learning is a technique that involves using a pre-trained machine
learning model as a starting point for training a new model. This can be particularly useful for
plant disease detection, as pre-trained models may have already learned features that are relevant
to the problem at hand.

5) Deployment considerations: Once a machine learning model has been trained for plant disease
detection, it must be deployed in a real-world setting. This may require integrating the model with
existing software systems, developing a user interface, and ensuring that the model is robust and
reliable.

48