Alzheimer Image Classification with CNNs 🧠

Welcome to the Alzheimer Image Classification repository! This project aims to provide an educational example of a multiclass image classification problem using convolutional neural networks (CNNs).

Project Overview 🔖

This repository contains the necessary code and guidance to build, train, and evaluate a CNN model that classifies brain scan images into four categories of dementia: non-demented, very mild, mild, and moderate dementia.

Learning Objectives 👨🏽‍💻

By working through this project, you will learn how to:

• Use torchvision for image processing tasks
• Create a .csv file mapping images to their labels
• Convert images to tensors and vice versa
• Build and train a CNN model for multiclass classification
• Implement a compartmentalized training loop for efficient model training and evaluation

Dataset 📁

The dataset used in this project is available on Kaggle. It includes brain scan images categorized into four classes:

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

Repository Contents 📦

This repository includes the following components:

• .csv annotations file builder (annotate.py)
• Dataset class: Handles data loading and preprocessing (dataset.py)
• Model definition: Defines the CNN architecture (model.py)
• Training script: Contains the training loop and model evaluation (train.py)

How To Get Annotations File: Example (CLI)

You can pass the following arguments to get the .csv file with the image annotations and labels to use in the __getitem__ method.

cd CNN4Alzheimer
python3 utils/annotate.py \
--root_dir ROOT_DIR_WITH_CLASS_SUBFOLDERS
--out_dir OUT_DIR
--col_names Images Labels