Pytorch implementation of differntiable orthogonal integral projection (backprojection) operator in Fourier space, based on our paper:
Differentiable probabilistic models of scientific imaging with the Fourier slice theorem (UAI 2019) Karen Ullrich, Rianne van den Berg, Marcus A. Brubaker, David Fleet, Max Welling
The requirements for the conda environment in which we have tested this code are started in requirements.txt
.
The main dependencies are
python 3.6
pytorch 1.1.0
A suitable conda environment may be installed via
conda env create -f requirements.yml
And used by
source activate backprojection
We provide a light introduction to scientific imaging in the jupyter-notebook cryo-tutorial.ipynb
. Specifically the generative model of scientific imaging observation_model.py
might prove useful for any application that involves orthogonal integral projection.
Please be warned that this repository is not going to be maintained regularly.
Please cite our paper if you use this code in your own work:
@inproceedings{ullrich2019backprojection,
title={Differentiable probabilistic models of scientific imaging with the Fourier slice theorem},
author={Karen Ullrich, Rianne van den Berg, Marcus A. Brubaker, David Fleet, Max Welling},
booktitle={proceedings of the Conference on Uncertainty in Artificial Intelligence (UAI)},
year={2010}
}