SuBongHyun / residual_diff_sparse

Residual-Guidance for Spase-View CT Reconstruction

Residual-Guidance Diffusion Model for Sparse-View CT Reconstuction

Environment

After conda install, follow the install.sh script. we provide a torch environment.

source install.sh

Then, install a torch-radon for fan-beam simulation. you can find other dependencies in requirements.txt.

Training

1. If you use your own dataset for the training, you can set your data directory at './configs/ve/AAPM_256_ncsnpp_continuous.py'

2. Run the following command for the train.

sh train.sh

Inference

1. If you use your own dataset for the sampling, please put them in './sample'

2. Run the following command for the test.

python run_CT_recon.py

Code is heavily based on score-mri and MCG.

We highly appreciate your support for sharing code.