Pytorch CNN model to diagnose Alzheimer's Disease on PET Scans
PyTorch version Ruben Aguilo Schuurs (UW-Madison), reworked from Dr. Alison Deatsch's Tensorflow Version
Files:
- main.py: Controls everything. Calls all of the following files
- utils/CNN.py & utils/CNN_Layers.py: CNN Model structure
- preprocess.py: prepares CustomDatasets training, validation, and testing
- train_methods.py: CNN training code
- Graphs.py: Multiple graphing / results functions
- K-fold.py: Specific graphs
- dataset_sd_mean_finder.py: unused
- show_image.py: show image
- TALOS_main.py: unsuccessful attempt at implementing talos
To run (in main.py):
1. Define the CNN filepath and dataset filepath
2. Change the properties
3. Comment or uncomment the functions (train, load, evaluate, predict) depending on your interest.
-- You don't need to load again after training.
4. See it in action! Training will take a while with large dataset!!