This is the codebase for my thesis, Conditional AnoDDPM.
The code is a fork Improved Diffusion, and implements Simplex noise as an alternative to Gaussian noise.
This section of the README walks through how to train and sample from a model.
Clone this repository and navigate to it in your terminal. Then run:
pip install -e .
This should install the canoddpm python package that the scripts depend on.
You can download the BraTS2020 dataset from here.
Run the datasets/brats.py script to pre-process the dataset.
The train_example.sh is an example script to train a model. It uses the scripts/multi_mri_train.py script.
Use the scripts/multi_mri_eval.py script to evaluate a model.