WSEAS TRANSACTIONS on COMPUTERS Ferdi Karakutuk, Ozer Ozdemir,

DOI: 10.37394/23205.2025.24.13 Asli Kaya Karakutuk

Plant Leaf Disease Detection using Machine Learning

AADITYA P. AGARKAR1, VIKRAMSINGH R. PARIHAR2

1, 2
Department of Electrical Engineering,
Prof Ram Meghe College of Engineering and Management,
Badnera-Amravati,
INDIA

Abstract: - In this research work, a prototype model based on machine learning is presented, which is
implemented as a Jupyter Notebook to predict harmful deviations from normal structures in plants or their
leaves. The model employs hybrid ensemble learning techniques in Python. The proposed system integrates
data obtained from 2,000 photos and applies machine learning algorithms to predict the crops that are best
adapted to the prevailing environmental conditions. This system aims to provide farmers with a variety of crop
options that are well-suited to their conditions and are presented in an informative and accessible manner.
Machine learning is essential for building computer programs that can learn and improve on their own by
analyzing data. Several models have been proposed for plant disease prediction, addressing some of the most
challenging issues in precision agriculture. Plant diseases are influenced by various factors, including fertilizer
usage, weather, soil variety, etc. This complexity requires working with multiple datasets. A prototype was
created using hybrid machine learning and data analytics with an algorithm in Python, designed to assist
farmers in managing plant health more effectively.

Key-Words: - Plant pathology, leaf disease, machine learning, CNN, feature extraction, disease classification,
precision agriculture.

Received: April 16, 2025. Revised: May 23, 2025. Accepted: June 15, 2025. Published: July 24, 2025.

1 Introduction actions so as to protect their crops. Ultimately, it not

Farmers are the backbone of the agricultural system, only helps the farmers by increasing their
and agriculture plays a key role in a country’s productivity but also the agriculture economy and
growth and development. In India, agriculture is not quality of yields.
only a major part of the economy but also provides
livelihoods for millions of people. Even so, there are 1.1 Problem Statement
many hindrances and challenges that the farmers As we already have seen that, plan diseases is a big
face, which inherently affects their yield and concern for farmers. We know that the early
revenue. detection of plant diseases is very necessary. To
One of the many challenges that a farmer faces improve the agricultural yield, it is important to
is the identification of a plant disease. Management monitor large crops and disease identification.
and treatment come hand in hand. Diseases damage
crops, which leads to less productivity and, In this research work, we aim to create a disease
ultimately financial losses. It has been observed that detection system that can classify various diseases
many farmers do not have enough resources to by analyzing the plant leaves. There are four main
detect these diseases early. Hence, they rely on stages in this system:
experience and or guesswork, which may lead to  First, a color transformation is applied to the
inaccuracy. input RGB image to prepare it for
Here, we can apply the usage of recent processing.
technologies like Artificial Intelligence and  The green pixels are removed by a
Machine Learning. Using machine learning, a threshold-based segmentation process
system can be created to detect diseases early and  The texture statistics are calculated for the
accurately based on the condition of leaves. These remaining useful part of the image
systems can analyze the leaves, flowers, and other  Finally, the classifier detects the various
parts of the plants and provide proper detection of features and accordingly classifies the
diseases. As a result, the farmers can take proper disease

E-ISSN: 2224-2872 136 Volume 24, 2025

 WSEAS TRANSACTIONS on COMPUTERS Ferdi Karakutuk, Ozer Ozdemir,
DOI: 10.37394/23205.2025.24.13 Asli Kaya Karakutuk

We use segmentation, which is threshold based, Crop diseases threaten food security, yet timely
in order to isolate the good and the infected parts of detection remains challenging due to limited
the leaves. We have used MATLAB simulation for infrastructure. This proposed approach [2] uses
programming and software development. This Random Forest (RF) to classify healthy and
system aims to provide farmers with a simple and diseased leaves, employing datasets with features
reliable tool to detect plant diseases early, helping extracted via Histogram of Oriented Gradients
them take timely action to protect their crops. (HOG). Machine learning on large datasets offers an
efficient solution for large-scale plant disease
1.2 Outline of The Project detection.
Various techniques have been employed for precise The demand for sustainable development has
and effective prediction of crops; out of which the advanced machine learning (ML) in agriculture,
Ensembling technique is used in our research. The transforming leaf disease classification in research
reason to use this is to enhance yield quality and and industry. This proposed approach [3] reviews
productivity. For high accuracy, the hybrid models datasets, ML techniques (traditional, deep, and
are used, and to do so, the data cleaning process augmented learning), and applications, offering
should be done effectively. The language used here resources to advance smart agriculture and leaf
is Python for the front-end as well as the backend. disease detection.
Agriculture, a key income source for many
1.3 Data Collection nations, faces financial losses from plant diseases
For this research, we have used a Kaggle data set. In caused by pathogens. It is very important for crop
this research, we have introduced a field data set. quality to effectively detect diseases from leaves.
This is done to diagnose and monitor its symptoms, Various techniques like computer vision, and
which is called a DioMOS plant. It is basically a machine learning AI based automatic disease
pilot data set. It contains the entire growing season detection are helping the farmers to detect the
of pear tree images. The duration for this is from diseases and act swiftly. This review paper [4]
February to July, which is a peak season to find the highlights these methods, describes various
best aspects of pear trees, for processes like techniques, and provides future directions in plant
classification and detection, various machine disease identification and classification.
learning and deep learning methods are employed. This proposed approach [5] presents an Ant
The images of some fruits and leaves are also Colony Optimization with CNN for plant disease
collected. The fruit is shown in various phases, such identification. This process works on the extraction
as fruit set, nut fruit, fruit growth, and ripening. In of various features like the color of leaves, texture
the same way, leaves are shown in various phases, of leaves, and live arrangement. The others claim
such as leaf spot, leaf, curl, slug, damage, and that their method outperforms existing techniques in
healthy leaf. accuracy.
In fuzzy sets, the various parameters like The plant diseases cause many losses, and
weather, air quality, agronomists, land quality, etc. manual detection is not reliable. Machine learning
are mentioned. Before creating the final rules, the and deep learning techniques are providing more
expert advice is taken into consideration. Also, the accurate solutions for plant disease identification.
various states of variables are taken into This review paper [6] describes the advancements in
consideration in the Fuzzy method. ml and DL techniques. The effectiveness of each of
these methods, their lacunas, and challenges are
mentioned. Also, the authors have provided their
2 Related Works reviews and recommendations to overcome these
challenges.
The leaf disorder identification approach focuses on
The traditional methods to detect crop diseases
detecting leaf diseases, analyzing research from
are accurate, whereas the AI-based technology
2010 to 2022. Techniques like multispectral
offers more accurate and reliable solutions. In this
imaging and ML models (e.g., SVM, Random
review paper, the various advancements [7] in AI-
Forest) and DL models (e.g., CNN, VGG, ResNet)
based plans this as detection techniques are
are widely used, with CNN, VGG, and ResNet
mentioned.
showing high efficacy, [1]. Performance is
A DL model using the YOLO technique is
evaluated using metrics like F1 score, precision, and
developed in this research work, [8]. The authors
accuracy. This review aids researchers in selecting
claim that they have achieved 99.8% accuracy in
efficient classifiers for leaf disease detection.
detecting plant leaf diseases, outdoing other

E-ISSN: 2224-2872 137 Volume 24, 2025

 WSEAS TRANSACTIONS on COMPUTERS Ferdi Karakutuk, Ozer Ozdemir,
DOI: 10.37394/23205.2025.24.13 Asli Kaya Karakutuk

classification models like CNN, KNN, and SVM. 4 Proposed System

The proposed approach focuses on blight, rust, and We will use the Hybrid ensemble technique for the
gray leaf spot, using single-leaf images. Future prediction of better crop yield [8]. The system uses
research should address environmental factors and only NPK (Primary Nutrients used for soil fertility,
multi-leaf disease detection. and various other factors are responsible for the
Using Convolutional Neural Networks (CNN) better growth of the crop. It will say which soil is
for early disease detection [9] offers an automated needed for the particular crop.
system to help farmers take timely action, reducing We have done an analysis to classify which
losses and boosting productivity. algorithm can be used for the soil. The flowchart in
Plant diseases threaten agriculture, but early Figure 2 represents the proposed machine learning
diagnosis using CNN-based models can improve algorithm for training the RF model.
food quality and reduce losses. This proposed
approach [10] fine-tuned pre-trained models like
DenseNet-121, ResNet-50, VGG-16, and Inception
V4 using the PlantVillage dataset, achieving 99.81%
accuracy, outperforming other models in
classification and performance metrics.

3 System Design
In fuzzy sets, the various parameters like weather,
air quality, agronomists, land quality, etc. are
mentioned. Before creating the final rules, the
expert advice is taken into consideration. Also, the
various states of variables are taken into
consideration in the Fuzzy method. A decision-
making framework of the existing system is as Fig. 2: Proposed machine learning algorithm for
shown in Figure 1. training the RF model

4.1 Exploratory Data Analysis

Another important entity for any data science
project is exploratory data analysis, which is often
overlooked or undervalued (EDA) [9]. We should
establish a link between the data that is used mostly
Fig. 1: Decision-making framework of Existing and rarely used data. The flowchart of data pre-
System processing in EDA is as shown in Figure 3.

3.1 Rule Set of Existing System

Frequently used in the control mechanisms of many
systems. Three steps had been taken to complete the
fuzzy logic operation:
1. Perplexity
2. Examining the rules
3. Defuzzification.

There are three types of graphical inference

techniques (fuzzy controllers) that can be used to
combine aggregation rules:
1. Mam-dani Systems
2. Sugeno Systems Fig. 3: Flowchart of Data Preprocessing in EDA
3. Trukamoto Systems.
It's often easier to compare the effects of
multiple exploratory techniques on each data
collection. This phase aims to understand the

E-ISSN: 2224-2872 138 Volume 24, 2025

 WSEAS TRANSACTIONS on COMPUTERS Ferdi Karakutuk, Ozer Ozdemir,
DOI: 10.37394/23205.2025.24.13 Asli Kaya Karakutuk

dataset, recognize missing values and outliers, if Feature variables: N, P, K, Temperature,

any, and get a sense of the story it tells using visual Humidity, pH, Rainfall Target variables: Label
and quantitative methods. It indicates the project's (Crop name)
next logical steps, concerns, or study areas. Figure 5 shows the Heat map for Correlation of
data with the features variables.

5 Result Analysis
Procedure
How do you run the project on the local host?
Open Anaconda In Prompt And Put Following
Commands
1. Cd..
2. Cd..
3. Cdleaf
4. Cdharvestify-master
5. Cdapp
6. Conda activates harvestify
7. [Link]

How to get recommended crop from this system

1. Open the crop recommendation page Fig. 5: Heat map for Correlation
2. Put all all the parameters from the data set
of crop recommendation From Figure 5, it is obvious that the feature and
3. Get the output target variables have a significant correlation. Thus,
there is a correlation between the feature and the
How to get recommended fertilizer for soil? target variable. And the data set is linearly
1. Open the fertilizer recommendation page separable. And also, it would fit perfectly for all the
2. Put all the parameters of soil from the adata linear machine learning models.
set of fertilizer recommendation
3. Get the output 5.2 Algorithms
Classification Vs. Regression
How To Get Plant Disease From Pictures Of Plant Predictive modeling problems involving
Leaf? classification vary from those involving regression.
1. Open the disease prediction page ● The task of classification is to predict a
2. Put the inflected pictures of the leaf discrete class mark.
3. Get the output of leaf disease predictive ● The role of regression is to predict a
(base) C:\leaf\Harvestify-master>pip install continuous quantity.
ipython Since we will be using one or more classes. So ,it is
coming under the classification predictive model
5.1 Correlation In Data Set:
Exploratory data analysis, a method of evaluating Accuracy Metrics
data sets in order to summarize the data's key Precision- The ratio of correctly predicted positive
characteristics, is as shown in Figure 4. observations to total predicted positive observations
is known as precision.
Precision=TP/TP+FP

Recall- The ratio of correctly expected positive

observations to all observations in the actual class is
called recall - yes.
Recall=TP/TP+FN

F1 score-The weight ever age of Precision and

Recall list the F1 Score.
F1Score=2*(Recall*Precision)/ (Recall+Precision)
Fig. 4: Data Summarization

E-ISSN: 2224-2872 139 Volume 24, 2025

 WSEAS TRANSACTIONS on COMPUTERS Ferdi Karakutuk, Ozer Ozdemir,
DOI: 10.37394/23205.2025.24.13 Asli Kaya Karakutuk

We have obtained 99.20% accuracy for our model 7 Future Scope

for better plant disease prediction. 1. Improved Model Architecture: Experiment with
different NN architectures or ensemble models to
Figure 6 represents the Graphical User Interface improve and increase yield, farmers' needs, and
(GUI) for our proposed model for plant leaf disease efficiency.
detection. 2. Enhanced Preprocessing Techniques: Explore
advanced preprocessing techniques to improve
model performance.
3. Real-time Feedback Mechanism: Implement a
real-time feedback mechanism based on detected
leaf disease to provide immediate responses or
interventions in applications such as mental
health monitoring or human-computer
interaction.
4. Multi-modal Emotion Detection: Integrate other
modalities, such as audio or text, to enhance leaf
disease detection accuracy and provide a more
comprehensive understanding of emotions.
Fig. 6: Graphical User Interface (GUI) of our 5. Emotion Recognition in Videos: Extend the
proposed model system to detect and recognize leaf disease
detection emotions in videos or live streams,
Output Screenshots: enabling applications in video analytics and
Figure 7 shows the Output of the model. Here, the surveillance.
information about the Late Blight disease in 6. Context-aware Emotion Detection: Develop
tomatoes is obtained when the query image of the context-aware models that consider contextual
plant leaf was given. information (e.g., user behavior, environment) to
improve leaf disease emotion detection and
interpretation.
7. Emotion-based Content Recommendation: Use
detected leaf disease to personalize content
recommendations in various applications such as
entertainment, marketing, and education.

References:
[1] Chittabarni Sarkar, Deepak Gupta, Umesh
Gupta, Barenya Bikash Hazarika, “Leaf
disease detection using machine learning and
Fig. 7: Output of the model deep learning: Review and challenges,”
Applied Soft Computing, Vol. 145, Issue 1,
2023, pp. 1-15,
6 Conclusion [Link]
In this project, ML algorithms are being used. We [2] Jianping Yao, Son N. Tran, Samantha Sawyer
have provided the classification whose dataset is & Saurabh Garg, “Machine learning for leaf
tremendously improved. So, this model too is able disease classification: data, techniques and
to accomplish with limited time constraints. It will applications”, Artificial Intelligence Review,
help in most agricultural practices. The losses Vol. 56, Issue 3, 2023, pp. 3571–3616,
should be less by using these techniques. There is [Link]
much future scope for other departments, such as [3] Shima Ramesh, Ramachandra Hebbar,
horticulture, and sericulture. Niveditha M., Pooja R., Prasad Bhat N.,
Shashank N., "Plant Disease Detection Using
Machine Learning," 2018 International
Conference on Design Innovations for 3Cs
Compute Communicate Control (ICDI3C),

E-ISSN: 2224-2872 140 Volume 24, 2025

 WSEAS TRANSACTIONS on COMPUTERS Ferdi Karakutuk, Ozer Ozdemir,
DOI: 10.37394/23205.2025.24.13 Asli Kaya Karakutuk

Bangalore, India, 2018, pp. 41-45, 12, Issue 10, 2022, pp. 2395-2413,
[Link] [Link]
[4] A. Bhargava, A. Shukla, O. P. Goswami, M.
H. Alsharif, P. Uthansakul, and M.
Uthansakul, "Plant Leaf Disease Detection, Contribution of Individual Authors to the
Classification, and Diagnosis Using Computer Creation of a Scientific Article (Ghostwriting
Vision and Artificial Intelligence: A Review," Policy)
IEEE Access, Vol. 12, Issue 1, 2024, pp. The authors equally contributed in the present
37443-37469, research, at all stages from the formulation of the
[Link] problem to the final findings and solution.
01.
[5] Yousef Methkal Abd Algani, Orlando Juan Sources of Funding for Research Presented in a
Marquez Caro, Liz Maribel Robladillo Bravo, Scientific Article or Scientific Article Itself
Chamandeep Kaur, Mohammed Saleh Al No funding was received for conducting this study.
Ansari, B. Kiran Bala, “Leaf disease
identification and classification using Conflict of Interest
optimized deep learning,” Measurement: The authors have no conflicts of interest to declare.
Sensors, Vol. 25, Issue 1, 2023, pp. 1-6,
[Link] Creative Commons Attribution License 4.0
[6] Muhammad Shoaib, Babar Shah, Shaker EI- (Attribution 4.0 International, CC BY 4.0)
Sappagh, Akhtar Ali, Asad Ullah, Fayadh This article is published under the terms of the
Alenezi, Tsanko Gechev, Tariq Hussain, and Creative Commons Attribution License 4.0
Farman Ali, “An advanced deep learning [Link]
models-based plant disease detection: A _US
review of recent research”, Frontiers in Plant
Science, Vol. 14, Issue 1, 2023, pp. 1-22,
[Link]
[7] Aria Dolatabadian, Ting Xiang Neik, Monica
F. Danilevicz, Shriprabha R. Upadhyaya,
Jacqueline Batley, David Edwards, “Image-
based crop disease detection using machine
learning,” Plant Pathology, Vol. 74, Issue 1,
2024, pp. 18-38,
[Link]
[8] Rosemary Ngozi Ariwa, Caleb Markus, Nora
Godwin Teneke, Shehu Adamu, and Fumlack
Kingsley George, “Plant Disease Detection
Using Yolo Machine Learning Approach,”
British Journal of Computer, Networking and
Information Technology, Vol. 7, Issue 2,
2024, pp. 115-129,
[Link]
EJWGFW6D.
[9] Sai Sharvesh R, Suresh Kumar K, and C. J.
Raman, “An Accurate Plant Disease Detection
Technique Using Machine Learning,” EAI
Endorsed Transactions on Internet of Things,
Vol. 10, Issue 1, 2024, pp. 1-9,
[Link]
[10] Andrew J., Jennifer Eunice, Daniela Elena
Popescu, M. Kalpana Chowdary, and Jude
Hemanth, “Deep Learning-Based Leaf
Disease Detection in Crops Using Images for
Agricultural Applications,” Agronomy, Vol.

E-ISSN: 2224-2872 141 Volume 24, 2025