plantdiseasedetectionsystem

Plant Disease Detection System

A deep learning-powered computer vision application for accurately detecting and classifying plant diseases from leaf images. This project leverages Convolutional Neural Networks (CNNs) and transfer learning techniques to help farmers, researchers, and agricultural experts monitor crop health and prevent yield losses.

---

Features

• Automatic Plant Disease Detection from leaf images.
• High Accuracy Model trained with advanced CNN architecture and fine-tuned hyperparameters.
• Real-time Prediction with support for both image uploads and live camera input.
• Google Colab + T4 GPU Compatible for fast training and inference.
• Model Export in .h5 format for easy deployment.
• User-Friendly Interface (optional integration with Streamlit/Flask for web app).

---

Project Structure

├── data/                  # Dataset (plant disease images)
├── notebooks/             # Jupyter notebooks for experiments
├── src/                   # Source code for training & evaluation
│   ├── data_loader.py
│   ├── model.py
│   ├── train.py
│   ├── evaluate.py
├── saved_models/          # Trained model (.h5 files)
├── requirements.txt       # Python dependencies
├── README.md              # Project documentation
└── app.py                 # (Optional) Deployment script

---

Installation

1. Clone the repository

git clone https://github.com/ayelefransi/plant-disease-detection.git
cd plant-disease-detection

1. Create virtual environment & install dependencies

python -m venv venv
source venv/bin/activate   # On Windows: venv\Scripts\activate
pip install -r requirements.txt

1. Run Jupyter Notebook (for training & testing)

jupyter notebook

---

Dataset

The dataset is sourced from PlantVillage and contains:

• Healthy leaves
• Leaves affected by fungal, bacterial, and viral infections
• Multiple plant species (tomato, potato, maize, etc.)

---

Model & Training

• Architecture: Custom CNN with fine-tuning using ResNet50 / EfficientNet.
• Optimizer: Adam with learning rate scheduling.
• Loss Function: Categorical Cross-Entropy.
• Hardware: Optimized for Google Colab T4 GPU.
• Output: Best-performing model saved as .h5 in Google Drive.

---

Usage

1. Training the Model

python src/train.py --data_path ./data --save_path ./saved_models/model.h5

2. Evaluating the Model

python src/evaluate.py --model ./saved_models/model.h5

3. Running the App (Optional Web Interface)

streamlit run app.py

---

Results

• Achieved accuracy: ~95%+ on test dataset.
• Model generalized well across multiple plant species.
• Reduced overfitting using data augmentation + dropout.

---

Future Improvements

• Expand dataset to cover more plant species.
• Improve real-time mobile deployment (TensorFlow Lite).
• Build a farmer-friendly mobile application.
• Integration with IoT for automated crop monitoring.

---

Contributing

Contributions are welcome! Please open an issue or submit a pull request.

---

License

This project is licensed under the MIT License.

---

Author

Fransi Ayele AI & Deep Learning Enthusiast | Agriculture + AI Researcher GitHub: ayelefransi

---