Classification of Rice Leaf Diseases Using Convolutional Neural Network (CNN)

Overview

This project implements a Convolutional Neural Network (CNN) model using TensorFlow/Keras to classify rice leaf images into six different disease classes.

The main objective of this project is to develop a deep learning model capable of identifying rice leaf conditions based on visual image characteristics.

The model performs image classification to distinguish healthy rice leaves and several common rice diseases.

---

Dataset

Dataset Information

Dataset used:

RiceLeafsv3

The dataset contains six rice leaf classes:

• bacterial_leaf_blight
• brown_spot
• healthy
• leaf_blast
• leaf_scald
• narrow_brown_spot

Dataset Distribution

Training data:

• 2,167 images

Validation data:

• 543 images

Due to the large dataset size, the original image dataset is not included in this repository.

---

Dataset Preprocessing

The image preprocessing steps include:

• Resize image into 128 × 128 pixels
• RGB normalization
• Data augmentation

Example visualization of dataset classes:

---

CNN Model Architecture

The CNN model was developed using TensorFlow/Keras with the following layers:

• Conv2D
• MaxPooling2D
• Flatten
• Dense
• Dropout

The model receives RGB images with size 128 × 128 × 3 and performs classification into six output classes.

Model architecture summary:

---

Training Configuration

The model was trained using the following configuration:

Parameter	Value
Optimizer	Adam
Loss Function	Sparse Categorical Crossentropy
Batch Size	32
Epoch	20
Input Image Size	128 × 128 pixels

---

Model Performance

The model evaluation results:

Metric	Result
Training Accuracy	85.83%
Validation Accuracy	86.00%
Average F1-score	0.85

---

Training Performance

The training process was monitored using accuracy and loss curves.

---

Model Evaluation

The model was evaluated using:

• Accuracy curve
• Loss curve
• Confusion Matrix
• Classification Report

Confusion matrix visualization:

The evaluation shows that the CNN model can classify most rice leaf images correctly according to their respective classes.

---

Prediction Result

The trained CNN model was tested using new rice leaf images.

The prediction output provides:

• Predicted class
• Confidence score

Example prediction results:

---

Project Structure

Mini_Project_CNN_RiceLeaf/

├── Mini_Project_CNN_RiceLeaf.ipynb

├── README.md

├── requirements.txt

├── LICENSE

└── assets/ ├── cnn_model_summary.png ├── confusion_matrix.png ├── prediction_result.png ├── riceleafsv3_classes.png └── training_history.png

Technologies Used

This project was developed using:

• Python
• TensorFlow
• Keras
• Jupyter Notebook
• Convolutional Neural Network (CNN)

---

How to Run the Project

Clone this repository:

git clone https://github.com/dedylesmana/Mini_Project_CNN_RiceLeaf.git

pip install tensorflow numpy matplotlib seaborn scikit-learn

Run the notebook:

Mini_Project_CNN_RiceLeaf.ipynb

Run all cells to reproduce:

• Dataset preprocessing
• CNN model training
• Model evaluation
• Rice leaf prediction

---

Conclusion

This project demonstrates the implementation of a Convolutional Neural Network (CNN) for rice leaf disease classification.

The model successfully classified six rice leaf classes with validation accuracy of approximately 86%.

The result shows that deep learning methods can be applied to support rice leaf disease identification based on image characteristics.

---

Author

M. Dedy Lesmana

Project: Rice Leaf Disease Classification Using Convolutional Neural Network (CNN)

Framework: TensorFlow & Keras

Method: Deep Learning - Convolutional Neural Network (CNN)