EyeAI-Retinopathy-Detection

🔍 Overview

EyeAI is an advanced deep learning system designed to detect and classify diabetic retinopathy from retinal images. Using state-of-the-art transfer learning and ensemble techniques, the system achieves high accuracy in identifying different stages of retinopathy, from mild to proliferative DR.

Key Features

• Multi-model ensemble learning with ResNet and VGG architectures
• Advanced image preprocessing techniques
• Explainable AI visualizations using GradCAM
• Comprehensive performance metrics including Cohen's Kappa

🏗️ Project Structure

EyeAI-Retinopathy-Detection/
├── bagging/
│   ├── resnet18Bagging.py
│   ├── resnet34Bagging.py
│   └── vgg16Bagging.py
├── boosting/
│   ├── resnet18Boosting.py
│   ├── resnet34Boosting.py
│   └── vgg16Boosting.py
├── stacking/
│   ├── resnet18Stacking.py
│   ├── resnet34Stacking.py
│   └── vgg16Stacking.py
├── aio.py
├── visualizations/
│   ├── gradcam/
│   └── training_metrics/
├── trained_models/
│   ├── resnet18/
│   ├── resnet34/
│   └── vgg16/
├── requirements.txt
└── README.md

🛠️ Installation

1. Clone the repository:

git clone https://github.com/yourusername/EyeAI-Retinopathy-Detection.git
cd EyeAI-Retinopathy-Detection

2. Create and activate virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

📊 Data Preparation

1. Download the DeepDRiD dataset from the provided source
2. Place the dataset in the appropriate directory:

mkdir data
mv downloaded_dataset data/deepdrid

🚀 Running the System

Quick Start with All-in-One Script

python aio.py

Running Specific Ensemble Methods

Bagging

cd bagging
python resnet18Bagging.py

Boosting

cd boosting
python resnet18Boosting.py

Stacking

cd stacking
python resnet18Stacking.py

🔧 Model Configuration

Available Models

• ResNet18
• ResNet34
• VGG16

Preprocessing Options

• Ben Graham Preprocessing
• Circle Cropping
• CLAHE
• Gaussian Blur
• Image Sharpening

Ensemble Techniques

• Bagging
• Boosting
• Stacking

📈 Performance Metrics

The system evaluates performance using:

• Cohen's Kappa Score
• Accuracy
• Loss Metrics
• ROC-AUC Curves

🎯 Results Visualization

Training Metrics

python visualizations/plot_metrics.py

Generate GradCAM Visualizations

python visualizations/generate_gradcam.py --model resnet18 --image_path path/to/image

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Commit changes
4. Push to the branch
5. Create a Pull Request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

📚 References

• DeepDRiD Challenge
• Diabetic Retinopathy Information
• GradCAM Paper

🙏 Acknowledgments

Special thanks to the DeepDRiD dataset creators and the medical professionals who contributed to the ground truth labeling.

📧 Contact

For questions or collaborations, please open an issue in the repository.