# Coffee Leaf Disease Classification Using Deep Learning 
## 📌 Introduction

This project introduces a robust AI-based system for classifying coffee leaf diseases using deep learning and hybrid classification models. The system is capable of detecting five types of leaf conditions: **miner, rust, healthy, phoma, and cercospora**. The main objective is to support early detection of diseases in coffee plants to enhance agricultural productivity and sustainability.

## 🧠 Models Used

The following models and hybrid configurations were implemented and analyzed:

- CNN1 (custom architecture)
- CNN2 (alternative custom CNN)
- VGG16 (transfer learning)
- CNN1 + SVM
- CNN1 + Random Forest
- CNN2 + SVM
- CNN2 + Random Forest
- VGG16 + SVM
- VGG16 + Random Forest
- ResNet50
- EfficientNetB0

## 📂 Dataset

- **Total images:** 58,555  
- **Categories:**  
  - `miner`  
  - `rust`  
  - `healthy`  
  - `phoma`  
  - `cercospora`  

Images were resized, normalized, and split into training and testing sets.

## 🧪 Preprocessing & Feature Extraction

To enhance feature representation and model accuracy, the following techniques were applied:

- **Data Augmentation:**
  - Horizontal Flip  
  - Vertical Flip  
  - Rotation (20°)  
  - Brightness Increase  
  - Gaussian Blur  

- **Texture-Based Feature Extraction:**
  - **LBP (Local Binary Pattern)**  
  - **XCS-LBP (Extended Center-Symmetric LBP)**  

These features were especially valuable for hybrid models like CNN + SVM or CNN + RF.

## 🧠 Model Explainability

To improve interpretability and transparency, the following techniques were integrated into the workflow:

- Grad-CAM  
- Grad-CAM++  
- SHAP  
- LIME  

These methods visually demonstrate how each model makes its decisions.

## 💻 Training Environment

- **Platform:** Kaggle  
- **GPU:** Tesla T4  
- **Framework:** TensorFlow 2.x / Keras  
- All models were trained using GPU acceleration

## 🚀 Project Structure

```
coffee_disease_classification/
├── data_preprocessing/
│   └── preprocess.py
├── feature_extraction/
│   ├── lbp_extraction.py
│   └── xcs_lbp_extraction.py
├── models/
│   ├── cnn1_model.py
│   ├── cnn2_model.py
│   ├── vgg16_model.py
│   ├── resnet50_model.py
│   ├── efficientnet_model.py
│   └── hybrid_models.py
├── evaluation/
│   ├── metrics.py
│   └── roc_curve_plotter.py
├── explanations/
│   ├── gradcam.py
│   ├── gradcampp.py
│   └── shap_lime.py
├── main.py
└── requirements.txt
```

## 🔧 How to Run

### 1. Clone the Repository

```bash
git clone https://github.com/your-repo/coffee-leaf-disease-classification.git
cd coffee-leaf-disease-classification
```

### 2. Install Dependencies

```bash
pip install -r requirements.txt
```

### 3. Prepare Dataset

Ensure the dataset is placed in the following directory structure:

```
dataset/
├── miner/
├── rust/
├── healthy/
├── phoma/
└── cercospora/
```

### 4. Preprocess and Extract Features

```bash
python data_preprocessing/preprocess.py
python feature_extraction/lbp_extraction.py
python feature_extraction/xcs_lbp_extraction.py
```

### 5. Train the Models

```bash
python models/cnn1_model.py
# Or try other models like:
# python models/vgg16_model.py
```

### 6. Evaluate Performance

```bash
python evaluation/metrics.py
python evaluation/roc_curve_plotter.py
```

### 7. Run Explainability Tools

```bash
python explanations/gradcam.py
python explanations/shap_lime.py
```