# Example expected training split paths (relative to the dataset root)
train/NonDemented/
train/VeryMildDemented/
train/MildDemented/
train/ModerateDemented/

# Download the dataset from Kaggle (link in README.md), unzip, and point the training script to the train/ directory.