Skip to content
Code for the article "Learning to solve inverse problems using Wasserstein loss"
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.gitignore
README.md
learned_primal_dual_l2.py
learned_primal_dual_wasserstein.py
phantom.py
test_wasserstein.py
wasserstein_util.py

README.md

Learning to solve inverse problems using Wasserstein loss

This repository contains the code for the article "Learning to solve inverse problems using Wasserstein loss".

Contents

The code contains the following

  • Training using circle phantoms

Pre-trained networks

The pre-trained networks are currently under finalization and will be released soon, in the meantime, training is just a few hours.

Dependencies

The code is currently based on the latest version of ODL and the utility library adler. They can be most easily installed by running

$ pip install https://github.com/odlgroup/odl/archive/master.zip
$ pip install https://github.com/adler-j/adler/archive/master.zip

The learning requires tensorflow, and the ray-transform needs ASTRA for computational feasibility

$ conda install -c astra-toolbox astra-toolbox

Authors

Jonas Adler, PhD student
KTH, Royal Institute of Technology
Elekta
jonasadl@kth.se

Axel Ringh, PhD student
KTH, Royal Institute of Technology
aringh@kth.se

Ozan Öktem, Associate Professor
KTH, Royal Institute of Technology
ozan@kth.se

Johan Karlsson, Associate Professor
KTH, Royal Institute of Technology
johan.karlsson@math.kth.se

Funding

Development is financially supported by the Swedish Foundation for Strategic Research as part of the project "Low complexity image reconstruction in medical imaging" and "3D reconstruction with simulated forward models".

Development has also been financed by Elekta.

You can’t perform that action at this time.