Benchmarking Deep Learning based CT Image Reconstruction Methods
Learned reconstruction is becoming more and more important for imaging tasks like computed tomography (CT) reconstruction. In this Code Sprint, we invite researchers and software developers to exchange knowledge, write code, form new collaborations and connect with industrial partners. As a concrete goal, we want to compare the performance of learned reconstruction methods in a challenge setup based on a medical and/or an industrial testing CT dataset.
See also the more detailed description (PDF).
The sprint was held on June 15–24, 2020 as a virtual event (due to the Covid-19 pandemic). In addition, we would like to organize a physical follow-up meeting in Bremen when the situation will allow it.
Challenges
The challenges (LoDoPaB-CT, Apples-CT) are still open. If you would like to join, please register as a participant.
Speakers
We were happy to have talks by the following speakers:
-
Prof. Dr. Peter Maass (University of Bremen, Germany)
Deep learning for CT image reconstruction
Slides -
Dr. Tatiana Bubba (University of Helsinki, Finland)
Deep neural networks for inverse problems with pseudodifferential operators: the case of limited angle tomography
Video, Slides, Paper -
Dr. Anna Trull (GREEFA b.v., Tricht, the Netherlands)
Deep learning & CT for fruit sorting and quality control -
Dr. Christian Etmann (University of Cambridge, UK)
iUNets: Fully-invertible U-Nets for Memory-Constrained Applications
Video, Slides, Paper