Deep learning for automated image analysis in generic imaging modalities

Prof. Dr. Thomas Brox (Department of Computer Sciences, University of Freiburg)

We will develop new deep learning methods in order to cover a broader set of tasks in biomedical data analysis. Concrete challenges to be approached will be the full applicability of deep learning to image sequences and volumetric images. We will also work on the challenge to obtain good results with little training data given for a particular task. To this end, we will investigate the relationship between deep networks trained for different tasks and a generic deep network trained on all of them jointly.