Fine-Tuning ResNet50 for Alzheimer's MRI Classification

This repository contains a Jupyter Notebook for fine-tuning a ResNet50 model to classify Alzheimer's disease stages from MRI images. The notebook uses PyTorch and the dataset is loaded from the Hugging Face Datasets library.

Table of Contents

• Introduction
• Dataset
• Model Architecture
• Setup
• Training
• Evaluation
• Usage
• Results
• Contributing
• License

Introduction

This notebook fine-tunes a pre-trained ResNet50 model to classify MRI images into one of four stages of Alzheimer's disease:

• Mild Demented
• Moderate Demented
• Non-Demented
• Very Mild Demented

Dataset

The dataset used is Falah/Alzheimer_MRI from the Hugging Face Datasets library. It consists of MRI images categorized into the four stages of Alzheimer's disease.

Model Architecture

The model architecture is based on ResNet50. The final fully connected layer is modified to output predictions for 4 classes.

Setup

To run the notebook locally, follow these steps:

1. Clone the repository:

   git clone https://github.com/your_username/alzheimer_mri_classification.git
   cd alzheimer_mri_classification
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

3. Open the notebook:

   jupyter notebook fine-tuning.ipynb
   

Training

The notebook includes sections for:

• Loading and preprocessing the dataset
• Defining the model architecture
• Setting up the training loop with a learning rate scheduler and optimizer
• Training the model for a specified number of epochs
• Saving the trained model weights

Example Training Code

# Training loop example
for epoch in range(num_epochs):
    model.train()
    running_loss = 0.0
    for images, labels in train_loader:
        images, labels = images.to(device), labels.to(device)
        optimizer.zero_grad()
        outputs = model(images)
        loss = criterion(outputs, labels)
        loss.backward()
        optimizer.step()
        running_loss += loss.item()
    print(f"Epoch [{epoch+1}/{num_epochs}], Loss: {running_loss/len(train_loader)}")