This repository contains the code for the article "Accelerated Forward-Backward Optimization using Deep Learning".

The code contains the following

• Training using anthropomorphic data from Mayo Clinic.
• Evaluation on example slice
• Implementation of the baseline methods.

The code depends on tensorflow-gpu=1.15.0, ODL and ASTRA toolbox. ODL and ASTRA can be installed by

$ git clone https://github.com/odlgroup/odl
$ cd odl
$ pip install --editable .
$ conda install -c astra-toolbox/label/dev astra-toolbox

Jevgenija Rudzusika,
KTH, Royal Institute of Technology
jevaks@kth.se

Sebastian Banert,
Lund University,
sebastian.banert@control.lth.se

Jonas Adler,
DeepMind,
jonasadler@google.com

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

Development is financially supported by the Swedish Foundation of StrategicResearch grant AM13-0049, grant from the VINNOVA Open Innovation Hubproject 2015-06759, and by Philips Healthcare.