Eye Disease Classification Using CNN, VGG16, and ResNet-50

Overview

This repository contains the implementation of an eye disease classification system using Convolutional Neural Networks (CNN), VGG16, and ResNet-50 architectures. The project aims to accurately classify various eye diseases from retinal images.

This project is part of the GirlScript Summer of Code 2024 program.

Table of Contents

• Overview
• Table of Contents
• Features
• Dataset
• Requirements
• Installation
• Usage
• Model Architecture
• Training
• License

Features

• Data preprocessing and augmentation.
• Implementation of CNN, VGG16, and ResNet-50 models.
• Training and validation scripts.

Dataset

The dataset used for this project consists of retinal images labeled with various eye diseases. You can download the dataset from Kaggle Eye Disease Dataset.

Requirements

• Python 3.7+
• TensorFlow 2.x
• Keras
• NumPy
• Pandas
• Matplotlib
• scikit-learn

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/eye-disease-classification.git
   cd eye-disease-classification

2. Install the required packages:

   pip install -r requirements.txt

Usage

1. Preprocess the Data:

   • Download the dataset and place it in the data/ directory.
   • Run the preprocessing script to prepare the data for training:

     python preprocess.py

2. Train the Model:

   • Train the CNN model:

     python train.py --model cnn

   • Train the VGG16 model:

     python train.py --model vgg16

   • Train the ResNet-50 model:

     python train.py --model resnet50

3. Evaluate the Model:

   • Evaluate the trained model:

     python evaluate.py --model model_name

Model Architecture

Convolutional Neural Network (CNN)

A custom CNN architecture designed for image classification.

VGG16

A pre-trained VGG16 model fine-tuned for eye disease classification.

ResNet-50

A pre-trained ResNet-50 model fine-tuned for eye disease classification.

Training

The training process involves:

• Splitting the dataset into training, validation, and test sets.
• Data augmentation to improve model generalization.
• Training the model with early stopping and model checkpointing.

License

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