A machine learning web application built with Python, OpenCV, scikit-learn, and Flask to detect whether a banana leaf is Healthy, Unhealthy, or Not a Leaf using image processing and KNN classification.

This project extracts texture and color-based features (using GLCM, LBP, and HOG) from banana leaf images, trains a K-Nearest Neighbors (KNN) classifier, and serves the prediction through a simple Flask web app.

• 🧠 Trained KNN model for 3-class classification
• 🎨 Uses advanced image feature extraction (GLCM, LBP, HOG)
• 🧾 Visualizes feature importance and class balance
• 🌐 Flask web interface for uploading and detecting banana leaves
• 🧩 Dataset augmentation and scaling with MinMaxScaler

project/
│
├── app.py                  # Flask app (main entry)
├── extract_features.py     # Feature extraction functions
├── generate_aug.py         # Data augmentation script
├── knn_trainer.py         # Model training script
├── scale.py               # Feature scaling and preprocessing
├── visualization.ipynb    # Data visualization and analysis
│
├── dataset/               # Dataset organization
│   ├── raw_data/         # Original dataset
│   │   ├── Diseased_leaf/
│   │   ├── Healthy_leaf/
│   │   └── Non_leaf/
│   ├── train_data/       # Training dataset
│   └── test_data/        # Testing dataset
│
├── static/               # Static files for web interface
│   └── styles.css        # CSS styling
│
├── templates/            # HTML templates
│   └── index.html       # Main web interface
│
├── uploads/             # Temporary storage for uploaded images
├── requirements.txt     # Python dependencies
└── README.md           # Project documentation

1. Clone the repository

   git clone https://github.com/Domincee/Banana-Leaf-Detector.git
   cd Banana-Leaf-Detector

2. Create a virtual environment (recommended)

   # Windows
   python -m venv venv
   .\venv\Scripts\Activate.ps1
   
   # Linux/MacOS
   python3 -m venv venv
   source venv/bin/activate

3. Install dependencies

   pip install -r requirements.txt

4. Run the Flask app

   python app.py

5. Open in your browser

   http://127.0.0.1:5000
   

Classification Report (Test Set)

            precision    recall  f1-score   support

Healthy Leaf 0.91 0.94 0.92 149 None-leaf 0.93 0.87 0.90 150 Unhealthy leaf 0.88 0.90 0.89 150

  accuracy                           0.90       449
 macro avg       0.90      0.90      0.90       449

weighted avg 0.90 0.90 0.90 449

The dataset consists of 2,000+ images, resized to 128×128, including:

• Healthy banana leaves (Augmented images)
• Diseased banana leaves (Actual images)
• Non-banana images (negative samples,self-collected)

⚠️ Raw and training datasets are not included in this repository due to file size limits. You can download them from: (https://drive.google.com/drive/folders/1mng06d0Y_U4hC7WM5hnbBNbuC5ohulcq?usp=sharing)

• Python 3.11
• OpenCV
• NumPy & Pandas
• scikit-learn
• scikit-image
• Matplotlib / Seaborn
• Flask

© 2025 Domince Aseberos. All rights reserved.

This project is released under the MIT License.

You are free to use, copy, modify, and distribute this software for educational or research purposes, provided that proper credit is given to the author.

⚠️ Note: The dataset and sample images are for demonstration and research purposes only.
They may contain content collected from public sources and are not included in this repository to comply with data-sharing and copyright policies.