PLANT DISEASE DETECTION USING

AI&ML
A PROJECT REPORT
Submitted by

CHANDRUGANTH E 721220205005

ROOPAN RAJ R 721220205043

RUBESH M 721220205044

SUNDAR S 721220205056

in partial fulfillment for the award of the degree

of

BACHELOR OF TECHNOLOGY

IN

INFORMATION TECHNOLOGY

KARPAGAM INSTITUTE OF TECHNOLOGY

ANNA UNIVERSITY : CHENNAI 600 025

MAY 2024
 ANNA UNIVERSITY : CHENNAI 600 025

BONAFIDE CERTIFICATE

Certified that this project report “PLANT DISEASE DETECTION USING

AI&ML” is the bonafide work of CHANDRUGANTH E (721220205005),
ROOPAN RAJ R (721220205043), RUBESH M(721220205044), SUNDAR S
(721220205056) who carried out the project work under my supervision.

SIGNATURE SIGNATURE

Dr. D. Bhanu., M.E., Ph.D., Mrs. M. SUDHA.,

HEAD OF THE DEPARTMENT SUPERVISOR

Professor, Assistant Professor,

Karpagam Institute of Technology Karpagam Institute of Technology

Department of Information Technology Department of Information Technology
Coimbatore-641105 Coimbatore - 641105

Submitted for the university project Viva-voce examination conducted at

Karpagam Institute of Technology, Coimbatore, on ……………

Internal Examiner External Examiner

I
 ACKNOWLEDGEMENT

With genuine humility, we are obediently thankful to God Almighty Without

him, this work would have never been a reality.

We express our profound gratitude to our respected Chairman

Dr. R. VASANTHAKUMAR, for giving this opportunity to pursue this course.
At this pleasing moment of having successfully completed the Project work.

We wish to acknowledge sincere gratitude and heartfelt thanks to our

respected Principal Dr. P. MANIMARAN M.E., Ph.D. for having us given the
adequate support and opportunity for completing the project work successfully.

We express our deep sense of gratitude and sincere thanks to our beloved
Vice Principal and Head of the Department Dr. D. BHANU M.E., Ph.D. who
has been a spark for enlightening our knowledge.

Our profound gratitude goes to our project coordinator

Mr. K. SRIRAM KUMAR M.E., (Ph.D.) and our project guide
Mrs. M. SUDHA and review members and all the faculty members of the
Department of Information Technology for the invaluable knowledge they have
imparted on us.

We express our gratitude to the HOD and all the staff members of the
Department of Information Technology for their support throughout the project.

Our humble gratitude and heartiest thanks goes to our family members and
friends for their encouragement and support throughout the course of this
project.

II
 ABSTRACT

The Plant Leaf Disease Detection Website is a critical challenges faced by the
agricultural sector in early detection and diagnosis of plant diseases.Using
Artificial Intelligence (AI) and Machine Learning (ML), this web-based
platform offers a user-friendly interface for farmers and agricultural experts to
identify and mitigate potential threats to crop yield. In which we are using
convolutional Neural Network for classifying Leaf images allowing it to
accurately classify and diagnose diseases based on visual symptoms . Once the
analysis is complete, the website provides a diagnosis of the disease affecting
the plant leaves. This includes information about the disease, its causes,
symptoms, and possible treatments. The processed images are then analyzed
using machine learning or deep learning algorithms trained to recognize various
plant diseases. The system compares the uploaded image with a database of
known disease images to make a diagnosis. Users can upload images of plant
leaves affected by diseases.The uploaded images are processed using image
processing techniques to identify patterns and anomalies that may indicate the
presence of disease.

III
 TABLE OF CONTENT

CHAPTER TITLE PAGE

NO. NO.
1 ABSTRACT III

INTRODUCTION

1.1 BACKGROUND 7

8
1.2 DEVELOPMENT OVERVIEW

2
BACKGROUND AND HISTORY
9
2.1 RELATED WORK

2.1.1 INTRODUCTION TO PLANT

9
DISEASE DETECTION
2.1.2 HISTORICAL METHODS AND EARLY 9
RESEARCH

2.1.3 TECHNOLOGY EVOLUTION IN 10

AGRICULTURE

2.1.4 RECENT DEVELOPMENTS IN 11

PLANT DISEASE DETECTION

5
2.2 TECHNICAL STRUCTURE 12

2.2.1IMAGE ACQUISITION 12

2.2.2 IMAGE PREPROCESSING 12

2.2.3 MACHINE LEARNING MODEL 13

2.2.4 MODEL EVALUATION 13

2.2.5 DEPLOYMENT 13

2.3 CONVOLUTIONAL NEURAL 13

NETWORK ARCHITECTURE

2.3.1 CONVOLUTION LAYER 14

2.3.2 POOLING LAYER 14

2.3.3 DENSE LAYER 14

2.3.4 OUTPUT LAYER 15

6
 CHAPTER 1

INTRODUCTION

1.1 BACKGROUND

The AI Engine trained based on the CNN Algorithm .The reason to use CNN
algorithm is it can automatically learn hierarchical representations of features
from the input data set given by the user.We can import the Pytorch package
from the cnn for Dynamic Computational Graph makes it easier to define and
modify complex CNN architectures Pytorch helps for the easy recognition of the
disease inthe plants based on the trained dataset CNN gives an accuracy rate of
more then 90% in disease detection and not affected by overfitting problem .
The first step in plant disease detection is to identify symptoms such as leaf
spots, wilting, discoloration, abnormal growth, or presence of pests. These
symptoms can vary depending on the type of disease and the plant species.
Managing data related to disease detection is important for tracking disease
prevalence, monitoring the effectiveness of treatment measures, and making
informed decisions about disease management strategies. Continuous
monitoring of plants is essential to detect new infections or disease outbreaks.
Preventive measures, such as crop rotation, use of disease-resistant varieties,
and proper sanitation practices, can help prevent future outbreaks . Overall,
plant disease detection is a complex process that requires a combination of
visual observation, laboratory analysis, and advanced technologies. Early

7
detection and effective management of plant diseases are essential for ensuring
healthy crops and sustainable agriculture.

1.2 DEVELOPMENT OVERVIEW

A plant disease detection website is an online platform

designed to help farmers, gardeners, and agricultural professionals
identify and manage diseases affecting plants. These websites use
technologies such as image processing and machine learning to
analyze images of diseased plants and provide accurate diagnosis and
treatment recommendations.. These websites use adva0nced
technologies such as image processing and machine learning to
analyze images of diseased plants and provide accurate diagnosis and
treatment recommendations. A plant disease detection website is an
online platform designed to help farmers, gardeners, and agricultural
professionals identify and manage diseases affecting plants. These
websites use technologies such as image processing and machine
learning to analyze images of diseased plants and provide accurate
diagnosis and treatment recommendations . The first step in plant
disease detection is to identify symptoms such as leaf spots, wilting,
discoloration, abnormal growth, or presence of pests. These symptoms
can vary depending on the type of disease and the plant species. If
visual inspection is not conclusive, plant samples may be collected for
further analysis. Samples may include leaves, stems, roots, or fruits
showing symptoms of disease. Image processing techniques and AI
algorithms are used to analyze images of plants and detect disease
symptoms. These technologies can quickly and accurately identify
diseases based on visual cues. Once the disease is identified,
8
 appropriate treatment measures can be implemented. This may include
the use of pesticides, fungicides, or cultural practices to manage the
disease and prevent its spread.
Sampling plant samples may be collected and sent to laboratories
for further analysis. Samples may include leaves, stems, roots, or fruits
showing symptoms of disease. Remote sensing technologies, such as
drones or satellites, are used to monitor crops for signs of disease.
These technologies can provide real-time data on plant health and
disease prevalence over large areas.

CHAPTER 2

BACKGROUND AND HISTORY

2.1 RELATED WORK

2.1.1 INTRODUCTION TO PLANT DISEASE DETECTION

Plant disease detection is crucial for maintaining the health and

productivity of plants, which is essential for food security and
environmental sustainability. Early detection of diseases can prevent
their spread and minimize damage to crops, leading to higher yields.
Accurate diagnosis helps in implementing timely and appropriate
management practices, such as using the right pesticides. By
identifying diseases early and applying targeted treatments, farmers
can reduce the need for broad-spectrum pesticides and reducing
environmental impact.
Protecting crops from diseases helps ensure a stable food supply,
which is essential for global food security. Farmers and agricultural
experts visually inspect plants for symptoms such as leaf spots,
9
 wilting, discoloration, or abnormal growth patterns. Agricultural
extension workers or researchers conduct field surveys to monitor the
prevalence of diseases in a region. Plant disease detection plays a
crucial role in ensuring food security, environmental sustainability,
and economic viability in agriculture. While traditional methods have
their limitations, advancements in technology, such as remote sensing,
image processing, and machine learning

2.1.2 HISTORICAL METHODS AND EARLY RESEARCH

Historically, farmers and gardeners relied on visual observation to
detect signs of plant diseases. They looked for symptoms such as
wilting, leaf spots, discoloration, abnormal growth, and presence of
pests. In the late 19th and early 20th centuries, scientists began to
study plant diseases more systematically. They developed techniques
to isolate and culture plant pathogens, laying the foundation for
modern plant pathology.
Microscopy became an important tool for plant disease detection.
Scientists used microscopes to examine plant tissues for signs of
disease, such as fungal spores, bacterial cells, or viral particles. Plant
Pathology The field of plant pathology emerged as a distinct discipline
in the early 20th century. Scientists like Anton de Bary and Heinrich
Anton de Bary made significant contributions to our understanding of
plant diseases and their causes.
Digital Imaging With the advent of digital imaging technology in the
late 20th century, researchers began to use cameras and image analysis
software to detect plant diseases. This laid the foundation for modern
computer vision-based disease detection systems
We create an AI model for detecting the plant disease identification
10
 using the CNN Algorithm in an efficient way
It overcome the problems in the existing system with the help of
CNN algorithm and identifies the disease including the ornamental
plants basically grown for aesthetics purpose
It provides suggestion of fertilizer needed to use for the disease based
on the land sector they own. The Additional features such as it
provides the certain sites for purchase the fertilizers and their contact
information

2.1.3 TECHNOLOGY EVOLUTION IN AGRICULTURE

Image processing techniques combined with machine learning

algorithms are used to analyze images of plant leaves and detect
disease symptoms. These systems can quickly and accurately identify
diseases based on visual cues. maintained by the blockchain and
therefore eliminate the need of a middleman.
Image processing techniques combined with machine learning
algorithms are used to analyze images of plant leaves and detect
disease symptoms. These systems can quickly and accurately identify
diseases based on visual cues.
Mobile applications are developed to help farmers diagnose plant
diseases, receive real-time disease alerts, and access information on
disease management practices.Near-Infrared Spectroscopy NIRS is
used to analyze the chemical composition of plant tissues. It can be
used to detect diseases based on changes in the chemical profile of
diseased plants compared to healthy ones.Hyperspectral imaging
involves capturing and analyzing a wide range of wavelengths of light
reflected by plants. This technology can detect subtle changes in plant
physiology that may indicate the presence of diseases .

11
2.1.4 RECENT DEVELOPMENTS IN PLANT DISEASE DETECTION
Convolutional Neural Networks CNNs are widely used for image
classification tasks, including plant disease detection. They are well-
suited for detecting patterns in images, making them ideal for
identifying disease symptoms in plant leaves . Deep learning
algorithms, which include CNNs and other neural network
architectures, are used to learn complex patterns in plant images and
accurately classify them as healthy or diseased.
Support Vector Machines SVM is a supervised machine learning
algorithm that can be used for classification tasks. It is often used in
conjunction with feature extraction techniques to classify plant images
based on disease symptoms.
Random forest is an ensemble learning algorithm that uses multiple
decision trees to classify data. It is used in plant disease detection for
its ability to handle large datasets and high-dimensional feature spaces
K-Nearest Neighbors KNN is a simple, instance-based learning
algorithm used for classification. It is used in plant disease detection to
classify plant images based on the similarity of their features to those
of known diseases.
Artificial Neural Networks ANNs are used in plant disease
detection for their ability to learn complex patterns in data. They are
often used in conjunction with other algorithms for feature extraction
and classification.
Transfer learning involves using a pre-trained neural network model
and fine-tuning it on a new dataset. It is commonly used in plant
disease detection to leverage the knowledge learned from large image
datasets.
Generative Adversarial Networks (GANs): GANs are used in plant
12
 disease detection for image augmentation and synthesis. They can
generate realistic images of diseased plants, which can be used to
augment training datasets.
Mobile applications are being developed to help farmers diagnose
plant diseases using their smartphones. These apps often use AI
algorithms to analyze images of diseased plants and provide treatment
recommendations.
3D imaging technologies are being used to create detailed models of
plants, which can be used to detect subtle changes in plant structure
that may indicate the presence of disease
Nanotechnology is being explored for the development of biosensors
capable of detecting plant pathogens at the molecular level. These
sensors could provide early detection of diseases before visible
symptoms appear.

2.2 TECHNICAL STRUCTURE

2.2.1 IMAGE ACQUISITION

Images of plants are captured using cameras or other imaging

devices. This can be done using smartphones, drones, or specialized
cameras designed for agricultural applications.

2.2.2 IMAGE PREPROCESSING

The captured images are preprocessed to enhance their quality and

remove noise. This may involve techniques such as resizing, color
correction, and noise reduction.

13
2.2.3 MACHINE LEARNING MODEL
A machine learning model, such as a convolutional neural network
(CNN) or a support vector machine (SVM), is trained using the
extracted features and labeled images of diseased and healthy plants.

2.2.4 MODEL EVALUATION

The trained model is evaluated using a separate set of images to
assess its performance in detecting plant diseases. This helps ensure
that the model can generalize well to new, unseen data.

2.2.5 DEPLOYMENT
Once the model is trained and evaluated, it is deployed to a production
environment where it can be used to analyze images of plants and detect
diseases in real-time.

2.3 CONVOLUTIONAL NEURAL NETWORK ARCHITECTURE

14
2.3.1 CONVOLUTION LAYER:
The convolution layer is a fundamental component of the CNN algorithm
responsible for extracting features from input images. In this layer, a set of
learnable filters or kernels are applied to the input images through a convolution
operation. Each filter acts as a feature detector, scanning across the input images
to detect patterns and spatial relationships. Through the process of convolution,
the filters extract features such as edges, textures, and shapes from the input
images, thereby capturing important information relevant to the classification of
plant diseases.

2.3.2 POOLING LAYER:

Following the convolution layer, the pooling layer plays a crucial role in
reducing the dimensionality of feature maps while preserving essential
information. Pooling operations, such as max pooling or average pooling, are
applied to the output of the convolution layer to downsample feature maps. By
aggregating neighboring pixel values, pooling layers help to abstract and
summarize the extracted features, making the representations more robust and
computationally efficient. This downsampling process also aids in reducing
overfitting and improving the model's ability to generalize to unseen data.

2.3.3 DENSE LAYER:

The dense layer, also known as the fully connected layer, serves as the final
stage of feature extraction in the CNN algorithm. In this layer, the flattened
output from the preceding layers is connected to a set of neurons, each of which
is associated with a weight parameter to learn. Through a series of matrix
multiplications and activation functions, the dense layer performs nonlinear
transformations on the extracted features, enabling the network to capture
complex relationships between different feature dimensions. This dense

15
representation forms the basis for making predictions about the presence or
absence of plant diseases.

2.3.4 OUTPUT LAYER :

The output layer is the final layer of the CNN algorithm responsible
for producing the network's predictions or classifications. Depending
on the specific task, such as binary classification (e.g., diseased vs.
healthy) or multi-class classification (e.g., identifying different types
of diseases), the output layer may consist of one or multiple neurons,
each corresponding to a distinct class label. Through the application of
appropriate activation functions, such as sigmoid for binary
classification or softmax for multi-class classification, the output layer
generates probability distributions over the possible classes, indicating
the likelihood of each class given the input images.

recommendation systems into the job search process holds promise for
enhancing user experience, increasing job satisfaction, and fostering
better matches between candidates and opportunities, ultimately
contributing to the advancement of both individuals' careers and
organizational success.

TRAIN ACCURACY :

16
INPUT

OUTPUT

17
RESULT :

A 95.6% accuracy rate was achieved using early stopping while

Training the model on 50 epochs.The Figure depicts the visualization
of training and validation accuracy. The result of detecting and
recognizing a strawberry plant is shown in Figure 8. On the left, a
healthy plant leaf, and on the right, a sick infected plant. The result of
detecting and recognizing a potato plant is shown in the Figure. On the
left, a healthy plant leaf, and on the right, a sick infected plant.

18