“LEAF CARE: A GUAVA LEAF DISEASE IDENTIFICATION

APPLICATION”

A PROJECT REPORT
Submitted to

ASSAM DON BOSCO UNIVERSITY

By
SHILPITA MEDHI
DC2019BTE0116
NIKHIL JAMES
DC2019BTE0106
AND
KUNAL KUMAR SHRIWASTAV
DC2019BTE0064

In partial fulfilment for the award of the degree

Of
BACHALOR OF TECHNOLOGY
IN
COMPUTER SCIENCE AND ENGINEERING

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

SCHOOL OF TECHNOLOGY
ASSAM DON BOSCO UNIVERSITY
AZARA, GUWAHATI 781 017,
ASSAM, INDIA.
BATCH (2019-2023)

i
 CERTIFICATE

This is to certify that the Project Report entitled “LEAFCARE: A GUAVA LEAF
DISEASE IDENTIFICATION APPLICATION” Submitted by SHILPITA
MEDHI(DC2019BTE0116), NIKHIL JAMES (DC2019BTE0106) and KUNAL KUMAR
SHRIWASTAV (DC2019BTE0064) to the Assam Don Bosco University, Guwahati, Assam,
in partial fulfilment of the requirement for Major project of 7th and 8th semester of Bachelor
of Computer Science and Engineering. It is a Bonafede record of the project work carried out
by them under my supervision during the year (2022-2023)

(Internal Guide Signature)

MRS. SMRITI PRIYA MEDHI

Assistant Professor (Senior)
Department of Computer Science and Engineering
School of Technology
Assam Don Bosco University, Azara
Date: ..........................................

ii
 CERTIFICATE

This is to certify that the Project Report entitled “LEAFCARE: A GUAVA LEAF
DISEASE IDENTIFICATION APPLICATION” Submitted by SHILPITA
MEDHI(DC2019BTE0116), NIKHIL JAMES (DC2019BTE0106) and KUNAL KUMAR
SHRIWASTAV (DC2019BTE0064) to the Assam Don Bosco University, Guwahati, Assam,
in partial fulfilment of the requirement for Major project of 7th and 8th semester of Bachelor
of Computer Science and Engineering. It is a Bonafede record of the project work carried out
by them under our supervision during the year (2022-2023)

Dr. Bobby Sharma Prof. Manoranjan kalita

Head of the department Director,School of Technology

Date:……………………. Date:………………………….

iii
 EXAMINATION CERTIFICATE

This is to certify that SHILPITA MEDHI, NIKHIL JAMES and KUNAL KUMAR
SHRIWASTAV bearing Roll Numbers DC2019BTE0116, DC2019BTE0106 and
DC2019BTE0064 respectively of the Department of Computer Science and Engineering
has carried out the project work in a manner satisfactory to warrant its acceptance and also
defended it successfully.

We wish them all the success in their future endeavors.

Examiners:

01. External Examiner:

02. Internal Examiner:

iv
 DECLARATION

We hereby declare that the project work entitled “LEAFCARE: A GUAVA LEAF
DISEASE IDENTIFICATION APPLICATION” submitted to the Assam Don Bosco
University, Guwahati, Assam, in partial fulfilment of the requirement for Major project of 7th
and 8th semester of Bachelor of Computer Science and Engineering. It is an original Work
done by us under the guidance of Mrs. Smriti Priya Mehdi (Asst. Professor, ADBU
Department of Computer Science and Engineering, School of Technology, Assam Don Bosco
University) and has not been submitted for the award of any degree.

(Signature of the student)

SHILPITA MEDHI
DC2019BTE0116
Department of Computer Science and engineering
School of Technology,
Assam Don Bosco University

(Signature of the student)

NIKHIL JAMES
DC2019BTE0106
Department of Computer Science and engineering
School of Technology,
Assam Don Bosco University

(Signature of the student)

KUNAL KUMAR SHRIWASTAV
DC2019BCA0064
Department of Computer Science and engineering
School of Technology,
Assam Don Bosco University

v
 ACKNOWLEDGEMENT

Words are not just enough to express our gratitude but we take this opportunity to express our
profound sense of gratitude and respect to all those who helped us throughout the duration of
this project.

We acknowledge the effort of those who have contributed significantly to our project. First of
all, we are very thankful to our parents for their regular support and guidance. We are also
very thankful to our college for providing us an opportunity to do this particular project
Special thanks to HOD of the Department of Computer Science and Engineering, Dr. Bobby
Sharma for letting us work on a project of this magnitude and providing assistance whenever
needed.

A very sincere gratitude to our project coordinator and our guide Mrs. Smriti Priya Medhi
for her constant support and guidance. We feel privileged to offer our sincere thanks and deep
sense of gratitude to her for expressing her confidence in us by his continuous support, help
and encouragement in implementing this project.

vi
 ABSTRACT

Diseases in plants cause major production and economic losses as well as reduction in both
quality and quantity of agricultural products. Now a day's plant diseases detection has received
increasing attention in monitoring large field of crops. Farmers experience great difficulties in
switching from one disease control policy to another. The naked eye observation of experts is
the traditional approach adopted in practice for detection and identification of plant diseases. In
this project we review the need of guava plant leaves disease identification system that would
facilitate advancements in agriculture. Early information on guava leaf health and disease
detection can facilitate the control of diseases through proper management strategies. This
technique will improve productivity of guava leaves. This project also compares the benefits
and limitations of these potential methods. It includes several steps image pre-processing,
features extraction, ResNet50, VGG 19

KEYWORD: Disease identification, Image pre-processing, classification, ResNet50,

VGG 19and conventional neural network.

vii
 LIST OF TABLES

Table _Title Page

3.1 Software Requirements (developers’ environment) 10

3.2 Hardware requirement (developers’ environment)

10

3.3 COCOMO Model Coefficient Values

13

viii
 LIST OF FIGURES

Table Title Page

1.1 DoctorP 17

1.2 Leaf doctor 18

3.1 Gantt chart for potato leaf detection 23

3.2 Work breakdown structure 24

4.1 Canker 26

4.2 Dot leaf 27

4.3 Healthy leaf 27

4.4 Mummification leaf 27

4.5 Rust leaf 27

4.6 ResNet50V2 28

4.7 Flow chart 30

4.8 Block Diagram 31

ix
5.1 Mount the dataset 32

5.2 Implementing the class 32

5.3 importing ResNet 33

5.4 model compilation 33

5.5 confusion matrix 34

5.6 User interface 35

5.7 front page implementation 35

x
 NOMENCLATURE

English Symbols

P precision

R Recall

TP True positive

TN True negative

FP False positive

FN False negative

xi
 ABBREVIATIONS

TFLITE TensorFlow Lite

CNN convolutional neural network

ResNet50V2 Residual Network

VGG Visual Geometry Group

xii
 CONTENTS

Title page………………………………………………………………………………….i

ACKNOWLEDGEMENT………………………………………………………………vii

ABSTRACT…………………………………………………………………………..….viii

LIST OF TABLES………………………………………………………………………..ix

LIST OF FIGURES………………………………………………………………….…...x

NOMENCLATURE………………………………………………………………….….xi

ABBREVIATIONS………………………………………………………………..…….xii

CHAPTER 1: INTRODUCTION

1.1 Title …………………………………………………………………………………...15

1.2 Objective……………………………………………………………………………... 15

1.3 Introduction……………………………………………………………………………15

1.4 Study of Existing System ……………………………………………………………..16

1.5 Motivation…………………… ……………………………………………………….18

CHAPTER 2: LITERATURE REVIEW

2.1 Literature review.………..…………………………………………………………….19

CHAPTER 3: REQUIREMENT ANALYSIS

3.1 System Requirements …………………………………………………………….......22

3.1.1 Software Requirements (for developer) …….……………………………...22

3.1.2 Hardware Requirements (for developer)………………………………...….22

3.2 Feasibility Study………………………………………………………………………23

3.2.1 Schedule Feasibility ………………………………………………………...23

3.2.1.1 Work Breakdown Structure …………………………………….....23

3.2.1.2 Gantt Chart ………………………………………………………..24

13
3.2.2 Economic Feasibility ……………………………………………………………...24

3.2.2.1 COCOMO Model ………………………………………….…....25

3.2.3 Operational Feasibility……………..………………………………………….......25

CHAPTER 4: PROPOSED PLAN

4.1 Description of dataset……………………………………………………………….....26

4.2 Methodology......…………………………………………………………………….....27

4.3 Flow chart……..….……………………………………………………………………30

4.4 Block diagram….……………………………………………………………………....31

CHAPTER 5: IMPLEMENTATIONS

5.1 dataset model implementation …………………………………………………………32

5.2 Users view…..………………..…………………………………………………………35

5.3 user interface implementation ………………………………………………………….36

CHAPTER 6: CONCLUSION AND FUTURE WORK

6.1 Summary……………………………………………………………………………….37

6.2 Future work…………………………………………………………………………….37

REFERENCES …………………………………………………………………………...38

CONFERENCE CERTIFICATE………………………………………………………..39

14
 CHAPTER 1

INTRODUCTION

1.1 TITLE: “LEAFCARE: AN GUAVA LEAF DISEASE IDENTIFICATION APPLICATION

1.2 OBJECTIVE:

1. To identify guava leaf diseases using Deep learning models and deep learningclassification
techniques.

2. To develop an application that showcases the identified disease and it’s associatedremedies
to the end user.

1.3 INTRODUCTION:

In this day and age, much of the labor is done by an app, such as a farming app, which will be extremely
helpful for many people who are new to farming, thus we believe that we should create an app to support
our agriculture sector. In many areas, the majority of homes have guava plants, and there may be some
disease in their leaves that people are unable to identify. The identification of leaf diseases is so critical
in agriculture. Leaf infections are unavoidable if proper safeguards are not followed in this field, and
they have massive consequences on plants, affecting quality, quantity, and productivity.

Psidium guajava, or delicious guavas, are native to Mexico, Central or South America, and the
Caribbean. It is high in vitamin C, calcium, pectin, and fiber. Because of challenges with natural and
environmental resources, technical issues, and other impediments, output is steadily diminishing. We
will, however, concentrate on the most critical challenges, such as illnesses that harm guava plants
and fruits, as well as early disease identification to prevent epidemics. Furthermore, the expert
system's early diagnosis of guava illness will result in improved harvests, assisting guava growers in
minimizing financial losses. As a result, image analysis proved to be an efficient technology that will
play an important part in the successful monitoring and identification of leaf disease situations.
Because visible patterns are accessible on plant leaves, patterns will be detected using various image
processing approaches to obtain a specific pattern relating to a disease, which will have an impact on
disease detection. As a result, the discovered features or patterns will be able to categories the disease
15
 when compared to past data, which can be accomplished using various machine learning approaches.
As a result, the combination of image processing and deep learning is particularly effective in disease
identification and classification.

To address this issue, we are first constructing a model, and then we will create an app that can detect
the disease using only a picture of a guava leaf and also provide treatment options. It is aimed not
only at farmers, but also at people who enjoy gardening at home. We will look for the following
guava leaf diseases: Mummification, Rust, Canker, Dot, and Wellness

1.3.1 What is Deep learning?

Deep learning is a machine learning technique that trains computers to do what people do
instinctively: learn by doing. Deep learning is a vital technology underpinning self-driving cars,
allowing them to recognize a stop sign or discriminate between a pedestrian and a lamppost. It is
essential for voice control in consumer electronics such as phones, tablets, televisions, and hands-
free speakers. Deep learning has received a lot of attention recently, and for good reason. It is
attaining results that were previously unthinkable.

A computer model learns to execute categorization tasks directly from images, text, or sound in deep
learning. Deep learning models can attain cutting-edge accuracy, sometimes outperforming humans.
Models are trained utilizing a vast quantity of labelled data and multi-layered neural network
architectures.

16
1.4 STUDY OF THE EXISTING SYSTEM
As part of our initial research, we decided to investigate applications that offer the same or similar
services.

1.4.1DoctorP
It is a crop disease detection app that detects plant leaf disease and is a research project. DoctorP is a
photo-identification programmed that can also snap a photograph of the unhealthy plant or select an
image from the device. It also supported the following crops: cucumbers, tomatoes, apples, grapes,
strawberries, wheat, corn, cotton, and after detecting the problem, it shows several ways to cure it.
Some screenshots of the app showed below –

Fig1.1: doctor p

1.4.2 Leaf doctor

Leaf doctor is an application that detects the proportion of healthy and diseased leaves. After
detecting the percentage of the leaves, the app saves the leaf history if the user desires, and the
user can also remove the history.
Some screenshots of the app showed below –

17
 Fig1.2: leaf doctor

1.5 MOTIVATION:
With the world's population growing, one of the most pressing challenges today is food production.
Food consumption is expected to have doubled by 2050. Food production requires a more yielding
and sustainable climate in order to increase plant yield. Guava was discovered in Portugal in the
early 17th century after being cultivated in the American tropics. It is popular in both tropical and
non-tropical countries, including Bangladesh, India, Pakistan, Brazil, and Cuba. Guava contains
phosphorus, calcium, nicotinic acid, and a variety of other essential nutrients. It also helps with
diabetes, regulates blood pressure, protects against dysentery, and treats diarrhea. Guava can grow
in a wide range of soil pHs (4.4 to 4.9), and it can tolerate significant and widespread climate change
in these soil types. It is the most cost-effective food and is regarded as a poor man's best friend.
Because guavas are so important in our Indian diets and are frequently consumed, it is also crucial
to keep a watch on their leaves while they are growing because they might become infected with
leaf illnesses. So, we are driven to create this app in order to identify these guava leaf diseases and
assist farmers in making the best option on how to deal with or cure the guava leaf disease.

18
 CHAPTER 2

LITERATURE REVIEW

2.1 LITERATURE REVIEW:

P. Perumala, Kandasamy Sellamuthub, K.Vanithac, and V.K.Manavalasundaramd et al. 2021 offer a

method for detecting guava leaf diseases early and precisely using image preparation methods and
Support Vector Machine (SVM). The proposed system includes the following stages: picture pre-
processing, image segmentation, image clustering using k-means, and extraction using Grey Level Co-
Occurrence Matrix (GLCM). After that, the image is classified using an SVM classifier. In comparison
to the existing framework, the suggested framework recognises plant leaf disease at an early stage and
improves accuracy to 98.17%.

Swarn Avinash kumar, Talha Meraj, Hafiz tayab Rauf et al. 2021 proposed a method for identifying
distinct guava plant species that uses a deep convolutional neural network (DCNN)-based data
improvement technique called colorhistogram equalisation and the unsharp masking technique. To
boost the amount of altered plant photos, nine angles from 360 were used. These enriched data were
then fed into cutting-edge classification networks. First, the proposed approach was normalised and
preprocessed. For the experimental evaluation, a locally gathered guava disease dataset from Pakistan
was employed. To detect diverse guava plant species, the proposed study employs five neural network
topologies, SqueezeNet and GoogLeNet. The experimental findings demonstrated that ResNet-101
achieved the maximum classification accuracy of 97.74%.

Ahmad Almadhor, Hafiz Tayyab Rauf 2, Muhammad Ikram Ullah Lali, and others proposed
innovations for 2021. To maintain high crop yields, automatic detection of illnesses in plants as they
appear on the plants' leaves and fruit is essential. In this paper, an artificial intelligence (AI)-driven
framework for detecting and classifying the most frequent guava plant diseases is given. The suggested
approach uses 4E colour difference image segmentation to separate disease-infected areas. Color
(RGB, HSV) histogram and textural (LBP) characteristics are also used to extract rich, useful feature
vectors. When compared to individual channels, the combination of color and textural data is utilized
to identify and achieve similar results, whereas illness recognition is accomplished by applying
advanced machine-learning classifiers (Fine KNN, Complex Tree, Boosted Tree, Bagged Tree, Cubic
19
SVM). On a high-resolution (18 MP) image dataset of guava leaves and fruit, the proposed framework
is tested. The Bagged Tree classifier produced the best recognition results on a collection of RGB,
HSV, and LBP features (95% accuracy in recognising four guava fruit diseases (Canker,
Mummification, Dot, and Rust) against healthy fruit). By adopting early precautions, the proposed
framework may assist farmers in avoiding potential production losses.

Farhan Al Haque, A S M; Rubaiya Hafiz, Md. Azizul Hakim, G. M. Rasiqul Islam, and others. 2020
published a paper in which they proposed a guava disease detection and cure suggestion offering
system based on convolutional neural networks (CNN). They gathered photos of guava infected by
anthracnose, fruit rot, and fruit canker, as well as disease-free guava, from several regions in
Bangladesh. In this work, they used three CNN models and discovered that the third model
outperformed the other two with an accuracy of 95.61%. Performance criteria such as precision,
recall, and F1 score are tested and determined to produce excellent outcomes via thorough
experimentation.

In 2016, Harshal Waghmare and Radha Kokare proposed solutions for plant disease identification
analysis and leaf texture pattern detection. This research is based on a method for detecting grapes
leaf disease. After context removal, the device is used as input for segmentation on a single plant leaf.
The sick component's image segmentation is then analysed using a high-pass filter for the leaf. It is
possible to obtain a unique sectional leaf texture. Nature invariant locally based fractal features
provide an excellent texture model. Each distinct sickness would have a different feel. The extracted
texture pattern is subsequently graded using multiclass SVM. Multiclass SVM implementations are
developed to diagnose diseases observed in grape plants for automated and farmer-friendly DSS
(Decision Support Systems). The Scheme segments and examines a single leaf, and the high pass
filter detects the sick area of the leaf. The strong texture module then provides a fractal-based retrieval
of segmented leaf texture function that is invariant locally in nature. The texture removed pattern is
then classed as an SVM designation for multiclass in groups that are either healthy or unwell. The
research focuses on two of the most common grass diseases: downy mildew and black red. The
recommendation approach quickly provides farmers with guidance from agricultural experts with 96.6
percent of accuracy.

20
Aditi singh and Harjit kaur 2021 proposed an image processing and machine learning-based technique for
diagnosing leaf image disease. The technique was provided for identifying and classifying potato plant
diseases. For this scenario, the Plant Village Dataset, which is well recognised as a freely accessible,
standard, and reputable data source, was evaluated. For image segmentation, the K-means strategy was
investigated, the grey level cooccurrence matrix concept was evaluated for feature extraction, and the
multi-class support vector machine methodology was considered for classification. The proposed
methodology was capable of obtaining an accuracy of 95.99 percent.

Jitesh P. Shah et al. (2016) study several image processing and machine learning strategies for detecting
sick plants in photographs. The document not only investigated several approaches, but it also included
key principles in image processing and machine learning for recognising and categorising plant diseases.
This research involved 19 document-based trials on several plant diseases affecting rice, fruit, and other
plants. The amount of image data collection, the absence of groups, preprocessing, technological
segmentation, styles, classifier accuracy, and other parameters were all considered. The survey was also
utilised to do additional study and improve the diagnosis and classification of a number of rice-related
disorders.

21
 CHAPTER 3

FEASIBILITY STUDY AND REQUIREMENT ANALYSIS

3.1 SYSTEM REQUIREMENTS (Technical feasibility):

3.1.1 Software Requirements (Development environment):

Table 3.1 Software requirement

SL NO. TYPE REQUIREMENT

1 FRONTEND XML 1.1

PROGRAMMING LANGUAGE Java 17, python 3.7.8

2

3 BROWSER Google Chrome

4 Operating System Windows 10 64 bit

5 Application Software/platform Android Studio 4.0, google Collab

3.1.2 Hardware requirements (developer environment):

Table 3.2Hardware requirement

SL NO. TYPE REQUIREMENT

1 Processor Intel i3 processor or above

2 RAM 4 GB preferred

3 HDD 500 GB

22
3.2 FEASIBILITY STUDY:

The feasibility study is an examination and analysis of the possibilities of a proposed project based
on comprehensive inquiry and research to support the decision-making process. A feasibility study,
also known as a feasibility analysis or feasibility report, is a method of determining whether or not
a project plan has a chance of success.

3.2.1 SCHEDULED FEASIBILITY

It is a measure of how realistic the project schedule is. It determines if a project can be completed
within the time range specified. This is demonstrated using the Work Breakdown Structure and
Gantt chart presented below.

3.2.1.1 GANTT CHART

Fig3.1 Gantt chart

23
 3.2.1.2WORK BREAKDOWN STRUCTURE

Fig3.2: work break down model

3.2.2 ECONOMIC FEASIBILITY

3.2.2.1 COCOMO MODEL

The basic COCOMO equation takes the form:

Effort applied (E) = a(KLOC)b [person-months]
Developed time (D) = c(Effort Applied)d
[months]
People required (P) = Effort Applied / Development Time [count],

24
where KLOC is estimated number of lines of code (expressed in thousands). The constants a, b, c and d are
given in the following table:
Table: 3.3 Constant values corresponding to software
1.Software project a b c d

2.Organic 2.4 1.05 2.5 0.38

3.Semi-detached 3.0 1.12 2.5 0.35

4.Embedded 3.6 1.2 2.5 0.35

Our project type is semi-detached project, Estimated LOC

=6000
Now, basic COCOMO equation of our project is: Effort
Applied (E) = a(KLOC)b [personmonths]
= 3.0 (6)1.12(person-months)

= 22 [person-months]
Development Time (D)
= c(Effort Applied)d
[months]
= 2.5(22)0.35[months] = [months]

People Required (P) = Effort Applied / Development Time [count]

= 22/7.35[count]=2.99 [count]

3.2.2.3 OPERATIONAL FEASIBILITY:

The operational feasibility of a proposed system is a measure of how well it solves issues, takes advantage
of opportunities found during scope definition, and satisfies the requirements identified during the
requirements analysis phase of system development. The suggested system is operationally practicable
because the project's ultimate output can be employed by consumers and will be simple and straightforward
to use. The mobile application in the proposed system does not require any specific training to operate. This
mobile application will meet the user's needs in emergency scenarios.

25
 CHAPTER 4
PROPOSED PLAN

4.1 DESCRIPTION OF DATASET

We have taken the dataset from Kaggle data. This dataset consists of 1500 image files of 5 different
classes, namely canker, dot, mummification, Rust, healthy.

The number of images is categorized as followed:

Table: 4.1. dataset description

class train test valid

canker 200 55 55

dot 200 55 55

Mummification 200 55 55

Rust 200 55 55

healthy 200 55 55

Total 1000(65%) 275(18%) 275(18%)

Fig 4.1:canker

26
 Fig 4.2: dot leaf

Fig 4.3: healthy leaf

Fig 4.4: mummification

Fig 4.5: rust leaf

4.2 METHODOLOGY FOR THE MODEL

Machine learning is becoming increasingly important in the automation of numerous processes.

With that goal in mind, the suggested framework was designed using deep learning approaches,
and we are currently in the process of detecting diseases from plant leaf photos. Each of these
processes is explained below.

Step-1: Image Pre-processing: This is a critical phase of the framework in which we rescale the
photos.

27
Step-2: CNN model design: Convolutional neural networks are increasingly being used as the deep-
learning technique for picture classification and recognition. The main advantage of employing
deep learning algorithms is that the picture feature extraction process is automated in deep learning.
CNNs are designed to map visual data (or 2D multidimensional data) to output variables (1-
dimensional data). It has been demonstrated to be so successful that they are ready to utilize it in
any type of data prediction, including image classifications.

Step-3: Training Data: The training data will be obtained from the processed data.

Step-4: Testing Data: The testing data will also be collected from the processed data, and it will
determine whether the disease is healthy or not.

Step-5: Evaluate performance.

4.2.1ResNet50V2

ResNet50V2 is a ResNet architectural version that is a deep convolutional neural network that is
often used for image classification applications. ResNet50V2 is an enhancement to the original
ResNet50 architecture, integrating additional strategies to improve training stability and accuracy
such as identity mapping, bottleneck blocks, and batch normalization.

4.2.2 VGG 19

VGG-19 is a 19-layer deep convolutional neural network. It is an enhancement to the previous

VGG16 architecture, with 19 layers of convolutional and pooling layers. As the name implies, the
VGG19 architecture has 19 layers. There are 16 convolutional layers and 3 completely linked layers
among the 19 layers.

4.2.3 PERFORMANCE EVALUATION MATRICS

a. Confusion matrix: A confusion matrix, also known as an error matrix in the field of machine
learning, is a specific table layout that allows visualization of the performance of an algorithm,
typically a supervised learning one (in unsupervised learning, it is usually called a matching matrix).

28
Each row of the matrix represents an actual class, whereas each column represents a predicted class,
or vice versa - both variations are available in the literature.

a. Accuracy score: One parameter for evaluating classification models is accuracy. Informally,
accuracy is the percentage of correct predictions made by our model. Formally, accuracy is
defined as follows:

For binary classification, accuracy can also be calculated in terms of positives and negatives as follows:

Where TP = True Positives, TN = True Negatives, FP = False Positives, and FN = False Negatives

b. F1 score: The F1-score combines a classifier's precision and recall into a single metric by taking
their harmonic mean. Its primary application is to compare the performance of two classifiers.
Assume classifier A has a higher recall but classifier B has a higher precision. The F1-scores
for both classifiers can be used to assess which one delivers superior results in this scenario.

The F1-score of a classification model is calculated as follows:

P = the precision

R= the recall of the classification model

Also, in the upcoming time we are thinking to go for some more techniques to develop the model.

29
4.3 FLOW CHIRT:

Fig4.7: Flow chart

30
4.4 BLOCK DIAGRAM

Fig4.8: Block diagram

31
 CHAPTER 5
IMPLEMENTATION
5.1 Dataset Model Implementation

5.1 mount the dataset

Description : Here first we import the necessary library and then mount the drive and provide the
image size and batch size.

5.2 implementing the classes

32
Description : Here we preprocessing the dataset and dividing the dataset and provide the image
size.

5.3 visualizing the data

Description: Here we have visualized the data and label the class

5.4 Importing ResNet50V2

33
5.5 Model compilation

5.6 confusion matrix

34
5.2 USER INTERFACE:

5.7 Users interface

5.3 APPLICATION DESIGN :

5.8 Front Page

35
 5.9 Disease identifying page

5.10 Image classifying code

36
 CHAPTER 6
CONCLUSION

6.1 SUMMARY

Guava leaf disease identification model using ResNet50 and VGG19 has showcased good
performance in guava leaf disease classification and has potential to impact guava farming
practices.in conclusion, the guava leaf disease identification model based on CNNs demonstrated a
significant advancement in disease diagnosis and management in guava farming. By early
identification the application has the potential to enhance crop productivity and reduces economic
loses. and contribute to the overall sustainability of guava cultivation.

6.2 FUTURE IMPLEMENTATION

1. In future we will try to add more guava leaf disease to our model.

2. Again for our application we will try to provide videos of treatment of diseases.

37
REFERENCES:

[1] Almadhor, A.; Rauf, H.T.; Lali, M.I.U.; Damaševičius, R.; Alouffi, B.; Alharbi, A. AIDriven
Framework for Recognition of Guava Plant Diseases through Machine Learning from DSLR Camera
Sensor Based High Resolution Imagery. Sensors 2021, 21, 3830

[2] H. Waghmare, R. Kokare, and Y. Dandawate, Detection and classification of diseases of Grape
plant using opposite color Local Binary Pattern feature and machine learning for automated Decision
Support System 2016, 3rd Int. Conf. Signal Process. Integ. Networks, 513– 8

[3] P. Perumala , Kandasamy Sellamuthub , K.Vanithac , and V.K.Manavalasundaramd,

classification of guava leaf disease using SVM Vol.12 No.7(2021),1177-1183

[4] Swarn Avinash Kumar Talha Meraj classification of guava leaf disease using deep
convulational Neural Network Appl. Sci. 2022, 12(1), 239

[5] Al Haque, A.F.; Hafiz, R.; Hakim, M.A.; Islam, G.R. A Computer Vision System for Guava
Disease Detection and Recommend Curative Solution Using Deep Learning Approach. In Proceedings
of the 2019 22nd International Conference on Computer and Information Technology (ICCIT), Dhaka,
Bangladesh, 18–20 December 2019; pp. 1–6

[6] Singh, A., & Kaur, H. (2021). Potato Plant Leaves Disease Detection and Classification using
Machine Learning Methodologies. In IOP Conference Series: Materials Science and Engineering (Vol.
1022, No. 1, p. 012121). IOP Publishing.

[7] J. P. Shah, H. B. Prajapati, and V. K. Dabhi, A survey on detection and classification of rice
plant diseases 2016, 2016 IEEE Int. Conf. Curr. Trends Adv. Compute.

38
 CONFERENCE

Paper will be published as a book chapter in Scopus Indexed Springer CCIS (Communication in Computer
and Information Science) Book Series (h-index)

39