The Home of Machine Learning at Leibniz University Hannover

  • Our work “Robust Shape Fitting for 3D Scene Abstraction” has been accepted for publication in the IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI).

The “Institut für Informationsverarbeitung” (Institute for Information Processing) is the home of machine learning at the Leibniz University Hannover. We focus on three main research directions, namely (I) computer vision & representation learning, (II) signal processing & -coding and (III) automated machine learning. Our methods range from deep learning, automated machine learning, reinforcement learning, image analysis, remote sensing and compression of audio, image, video as well as DNA to biomedical data. Our efforts are directed towards making efficient use of multi-modal and high dimensional data for reliable predictions, ultimately supporting end-users, developers and decision makers in a vast range of applications.

Since the foundation in 1973, the institute holds a strong tradition of cultivating connections to industry partners and jointly developing solutions to automatically process and harness information. Some of the developed methodology were successfully spun-off commercially as for instance with driver assistance modules, cochlear implants or component testing. The institute is also well known for being actively involved in the standardization of MP3, MPEG-2, AVC (H.264), HEVC (H.265) as well as MPEG-G.

Do you want to join us? We have open positions.

Current Spotlight
Mobile Computer Vision
Computer vision on mobile devices, e.g. Apple's iPhone and Android devices (more...)

DAC: Dynamic Algorithm Configuration
Algorithm Configuration (AC) aims to optimize algorithm performance by automating important decisions like hyperparameter settings and algorithm choice. Both theoretical and empirical results have shown, however, that only making these decisions once at the beginning of an algorithm run is often not optimal. Instead, the best configuration often depends on the current timestep and algorithm's state. Therefore Dynamic Algorithm Configuration (DAC) learn configuration schedules that fit the current state to improve overall performance. (more...)


TNT
"Am tnt erhalte ich für Projektarbeit & Lehre sehr viel Vertrauen, kann eigene Vorschläge einbringen und Forschung & Lehre auf diese Weise aktiv mitgestalten. Im Rahmen der Forschung genieße ich viel Gestaltungsspielraum und schätze die vielseitigen Themen hier am Institut, welche einem viele Blicke über den Tellerrand ermöglichen. Als Familienvater schätze ich außerdem die flexiblen Arbeitszeiten und das damit verbundene Verständnis."
Alexander Lange