Image Analysis and Learning

Learning-based Image Analysis for Industry and the Life Sciences

Vast numbers of images are now routinely acquired to explore complex spatially organized processes in areas as diverse as biology and industrial quality control. We develop algorithms to solve real-world image analysis problems from these domains. Beyond conventional images, we often deal with data where a multitude of observations are available for each element, such as multidimensional or spectral images.

Whenever reasonable, we use techniques from statistical or machine learning to build such automated diagnostic systems. Our ambition is to make the machine learning process as robust and user-friendly as possible: we hence develop algorithms for active and semi-supervised learning that can benefit even from small or partially annotated training sets while exploiting spatial context. We seek to make state-of-the-art algorithms available to practitioners in the guise of ilastik.

We enjoy and cultivate very close collaboration with our experimental partners from biology and engineering and are grateful for the constant flow of challenging problems and constructive criticism they provide. A long-standing partner and sponsor of this group is the Robert Bosch GmbH, for whom we have developed and implemented a number of systems that successfully run in a production environment, 24/7 and all year round.


  • 28.01.2016: We are now on top of the Connectomics Challenge Leaderboards both for the ISBI 2012 and the SNEMI 3D challenges. Congratulations to the whole team, and Constantin Pape in particular!
  • Anna Kreshuk will advise her first own PhD students thanks to funding she won from the Baden-Wurttemberg "Elite PostDoc program". Congratulations!
  • Jens Kleesiek, Gregor Urban and others win the 1st and 3rd prize at the MICCAI BraTS (Brain Tumor Segmentation) Challenge.
  • Two orals at CVPR 2014
  • Alumnus Bjoern Menze now Assistant Professor at TU München
  • Papers at NIPS, CVPR (x3), ICML, ECCV (x3), UAI, ICPR
  • Xinghua Lou and Luca Fiaschi have won the best paper award at the MICCAI-Workshop on Machine Learning in Medical Imaging. Congratulations!