Plant Disease Detection

using Machine Learning

Course Code : ECP 402
Presented By : Akash (211220003)
Hitaishi Goel (211220021)
Supervised By : Dr. Sandeep Kumar
Presented to : Dr. Nitin Singh Singha
Dr. Mahesh Kumar Singh
Introduction
Threat to Early Detection is
Agriculture Crucial
Plant diseases pose a Early detection enables timely
significant threat to global intervention, preventing the
food security, leading to crop spread of diseases and
yield losses and economic minimizing damage.
hardship.

Limitations of Traditional Machine Learning and

Methods Computer Vision
Traditional disease detection ML and Computer Vision offer

methods are often manual, automated, accurate, and

time-consuming, and efficient solutions for plant

subjective, hindering effective disease detection.

management.
Literature Review
1 SVM-based 2 Deep Learning with 3 Base Paper: Custom
Classification VGG16 CNN
Proposed a custom CNN model
Utilized Support Vector Employed deep learning with for identifying diseases in
Machines (SVM) to classify a pre-trained VGG16 model for various crops, achieving 90%
diseases in tomato leaves, plant disease detection, accuracy on the Plant Village
achieving 80% accuracy. achieving 85% accuracy. Dataset.
Research Gap
Limitations of Existing Focus on Lightweight Multi-class Classification
Methods Architectures
Most methods rely on pre-trained Limited research on lightweight, custom Insufficient exploration of multi-class

models, which may not generalize well architectures that achieve high accuracy classification for both common and rare

to new data. with lower computational overhead. plant diseases.

 Objectives
1 Design a Custom 2 Evaluate
CNN Performance
To develop a custom To evaluate the performance
Convolutional Neural of the proposed model on
Network (CNN) for accurate different classes like
and efficient detection of apple_blackrot,
multiple plant diseases. apple_healthy,
tomato_healthy and more.

3 Comparison with Existing

Techniques
 Methodology
1 Data Collection &
Preprocessing
Resize images, normalize pixel values, and augment data for better generalization.

2 Model Development
Design a custom CNN with convolutional layers for feature extraction, pooling layers
for dimensionality reduction, and fully connected layers for classification.

3 Training and
Evaluation
Train the model using cross-entropy loss function and Adam optimizer, evaluating
performance using accuracy, precision, recall, and F1-score.

4 Improving & Comparison

Create a user-friendly interface for real-time disease detection based on uploaded
leaf images.

5 Documentation
Document the entire process for reproducibility and future enhancements.
Data Data
Collection
1 Source
The dataset used in this project
Preprocessing
1 Image Resizing
Images are resized to a uniform size to
was obtained from the Plant
ensure consistency during model training.
Village Dataset.

2 Data
Augmentation
Artificial data is generated through techniques
2 Content
like rotation, flipping, and color shifting to
The dataset contains images of
increase the size and diversity of the training
healthy and diseased leaves from
dataset.
six classes. 3 Labeling
Each image is meticulously labeled to
3 Classes indicate whether it represents a healthy or
The six classes include: apple black diseased plant.
rot, apple healthy, tomato healthy,
tomato late blight, potato healthy,
and potato late blight.
Custom CNN
Model
Development
Model Training Model Evaluation
Architecture: We developed a custom Convolutional Neural Network (CNN) Metrics Used:
architecture with the following layers:
● Accuracy: The ratio of correct predictions to the total predictions.
● Convolutional Layers: Extract features from input images using filters. ● Precision: The ability of the model to return relevant results for
● Pooling Layers: Reduce the spatial dimensions of the features, each class.
retaining essential information. ● Recall: The model’s ability to detect all the relevant cases in each
● Fully Connected Layers: Perform the final classification into one of the class.
six classes. ● F1-Score: A balance between precision and recall, providing a
single metric for model performance.
Loss Function:
Confusion Matrix:
● We used the Cross-Entropy Loss function, which is well-suited for multi-
class classification tasks. ● Used to visualize the model’s performance across the different
classes, showing true positives, false positives, false negatives, and
Optimizer: true negatives.
● The model was trained using the Adam optimizer, which adapts the Test Accuracy:
learning rate during training, making it efficient in handling sparse
gradients. ● The model achieved a test accuracy of 92%, demonstrating its
effectiveness in classifying plant diseases across the six classes.
Batch Size and Epochs:
Validation:
● A batch size of 32 and a total of 50 epochs were used, allowing the
model to learn effectively without overfitting. ● The model was validated using unseen data, and the results were
compared with those of the training process to ensure no overfitting
Data Augmentation: occurred.
● Real-time data augmentation was applied during training, helping the
model generalize better by exposing it to various transformations of the
input images.
Model Performance Overview
Documentation
Code Documentation: Clear
explanations for data preprocessing,
CNN architecture, and training, with
modular code and comments.

User Guide: Instructions on using

the interface, uploading images, and
troubleshooting common issues.

Technical Documentation: Details

system architecture, performance
metrics, and hardware requirements
for easy future updates.
Results and
Discussion
Dataset Plant Village Dataset

Classes apple_blackrot,
apple_healthy,
tomato_healthy,
tomato_late_blight,
potato_healthy,
potato_late_blight
Accuracy 96%

Precision 95%

Recall 95%

F1-Score 95%
Conclusion and Future
Work
Conclusion Future Work Impact
We plan to extend the model to
A custom CNN model was successfully This research contributes to the
detect a wider range of plant diseases
developed but implemented only for and explore transfer learning development of intelligent systems for
detection of plant diseases for 6 techniques to improve performance enhancing plant disease management,
with limited data. Additionally, We
classes for now and achieved aim to implement disease remedy ultimately improving agricultural
significant improvements compared to suggestions alongside the disease productivity and food security.
predictions.
existing methods.
References
1. P. K. V, E. G. Rao, G. Anitha and G. K. Kumar, "Plant Disease Detection using Convolutional Neural Networks," 2021 5th
International Conference on Trends in Electronics and Informatics (ICOEI), Tirunelveli, India, 2021.
2. S. Duth P and J. T. N, "Tomato Fruits Disease Detection Using Image Processing," 2022 International Conference on Futuristic
Technologies (INCOFT) Karnataka, India, 2022.
3. A. Kiran, S. K. L. Naik, M. S. Raj, and S. K. Palvadi, " Plant Disease Detection using Image Processing with Machine Learning,"
Proceedings of the Fourth International Conference on Electronics and Sustainable Communication Systems (ICESC-2023) IEEE
Xplore.
4. M. S., K. R. R., S. K., and P. G., " Disease Detection in Fruits using Image Processing," Proceedings of the Sixth International
Conference on Inventive Computation Technologies (ICICT 2021), Erode, India, 2021.
5. S. U. Park, H. Kim, and C. Lee, "Plant Leaf Disease Detection using a Convolutional Neural Network and Data Augmentation,"
2019 21st International Conference on Advanced Communication Technology (ICACT), PyeongChang, Korea (South), 2019.
6. M. Singh and M. Kaur, "Classification of Crops Disease by Using Convolutional Neural Networks (CNN)," 2020 6th International
Conference on Advanced Computing and Communication Systems (ICACCS), Coimbatore, India, 2020.
7. J. P. Singh, R. K. Sahu, and S. D. Rautaray, "Plant Disease Detection Using Machine Learning," 2016 IEEE 8th International
Conference on Computational Intelligence and Communication Networks (CICN), Tehri, India, 2016.
8. K. S. Jadhav, A. D. Patil, and B. P. Patil, "Identification of Plant Diseases Using Convolutional Neural Networks," 2020
International Conference on Emerging Smart Computing and Informatics (ESCI), Pune, India, 2020.
9. K. Subramanian and K. D. Patel, "Deep Learning Approaches for Plant Disease Detection Using Leaf Images," 2018 International
Conference on Advances in Computing, Communications and Informatics (ICACCI), Bangalore, India, 2018.