/PnP-MonteCarlo

Published in IEEE Trans. Comput. Imag.

Primary LanguagePython

Plug-and-Play Monte Carlo

This is the code repo for the paper "Provable Probabilistic Imaging using Score-based Generative Priors" (IEEE TCI | arXiv | webpage).

Environment

Setup the environment.

conda env create --file pmc.yml

Activate the environment.

conda activate pmc

Deactivate the environment.

conda deactivate

Checkpoints

The pre-trained checkpoints are available here. Note that the checkpoints need to be placed in the proper fold for loading.

Create the folder

mkdir score_ckpt

Move the file to the folder

mv dir/to/ckpt ./score_ckpt

Run

To run the code, simply type

python run_pmc.py -c dir/to/config.yml

Please add or configurate the data loader (pmc/test_datasets/) to allow proper loading of your own data.

Citation

@ARTICLE{10645293,
  author={Sun, Yu and Wu, Zihui and Chen, Yifan and Feng, Berthy T. and Bouman, Katherine L.},
  journal={IEEE Transactions on Computational Imaging}, 
  title={Provable Probabilistic Imaging Using Score-Based Generative Priors}, 
  year={2024},
  volume={10},
  number={},
  pages={1290-1305},
  doi={10.1109/TCI.2024.3449114}}