Differentiable Uncalibrated Imaging

This repository contains the code for Differentiable Uncalibrated Imaging. The paper is available at https://arxiv.org/abs/2211.10525.

Python setup

This code was developed with Python 3.9.12. Run the following code to install the conda environment provided in environment.yml:

conda env create -f environment.yml

There may be some pip failures which is alright. Package ml-collections may have to be explicitly installed using pip. Install it using pip of the newly created environment:

<PATH_TO_ANACONDA3_INSTALLATION>/envs/diff_imaging/bin/pip3 install ml-collections

where <PATH_TO_ANACONDA3_INSTALLATION> should be replaced with the path to your Anaconda 3 installation such as

~/anaconda3/envs/diff_imaging/bin/pip3 install ml-collections

Similarly install odl:

<PATH_TO_ANACONDA3_INSTALLATION>/envs/diff_imaging/bin/pip3 install git+https://github.com/odlgroup/odl.git@0b088df8dc4621c68b9414c3deff9127f4c4f11d

Getting trained Unet models and data

Download trained Unets and the data from here to the home directory of this repository.

Solving inverse problems

Solve inverse problems by running the code:

mkdir work_directory
 python main.py --config=configs/<CONFIG_FILE> --config.workdir=work_directory

<CONFIG_FILE> should be replaced by the configuration file of the type of experiment you wish to run. These files are in the configs/ directory and contain the experiment parameters:

• ct_2d_implicit_neural_config.py: 2D CT experiments using implicit neural measurement representations
• ct_2d_spline_config.py: 2D CT experiments using spline measurement representations
• ct_3d_implicit_neural_config.py: 3D CT experiments using implicit neural measurement representations
• ct_3d_spline_config.py: 3D CT experiments using spline measurement representations These configuration files will use the downloaded trained Unets and data.

Citing this work

If you find this code or the paper useful in your work, please consider citing the paper:

@article{gupta2022differentiable,
  title={Differentiable Uncalibrated Imaging},
  author={Gupta, Sidharth and Kothari, Konik and Debarnot, Valentin and Dokmani{\'c}, Ivan},
  journal={arXiv preprint arXiv:2211.10525},
  year={2022}
}