This is the code which incorporate Noise2Noise network fine-tuning into iterative reconstruction at testing time.
- Miniconda, the environment is given in env.yml
- https://github.com/wudufan/ReconNet for the reconstruction backend. We are mitigating everything to a new framework https://github.com/wudufan/CTProjector, but please use the old repo at the moment.
./preprocessing provides codes to generate training/testing dataset from the rebinned projection data from 2016 Low-dose CT Challenge.
./eval provides evaluation and plotting scripts.
Each folder gives the reconstruction scripts for the corresponding algorithm.
The Script folder under each gives the bash and slurm scripts to call the python scripts.