Multimedia Information Processing Group


  • Experitur - Automating Machine Learning


    Experitur automates machine learning and other computer science experiments and stores the results in an easily accessible format. It includes grid search, parameter substitution and resuming aborted experiments.

    The tool as well as instructions on the usage can be found in this GitHub repository.

    Contact/Maintainer: Simon-Martin Schröder


  • µCT Toolbox


    A MATLAB toolbox for working with very high resolution (e.g. micro-)CT volumes.

    • 3D cortex segmentation
    • Cortical Thickness measurement
    • Mesh creation for cortical surface

     

    Example output

    Example segmentation Meshing example output
    Central slice of segmentation result Mesh of cortical center surface colored with Ct.Th.

     

    Binary releases are available for Linux, Windows and macOS.

    The repository can be found here.

    Contact/Maintainer: Stefan Reinhold


  • Cortex Identification in Quantitative CT


    Identifying the conter of the cortical bone in clinical CT scans is a tedious task. Furthermore, the apparent cortex center may be displaced from the real cortex center, making it impossible for a human operator to precisely identity it. This software automates the process of cortex center identification and has proved to have very high precision in an ex-vivo setting.

    The software library CortidQCT is based on the publication "Reinhold S. et al. (2019) An Analysis by Synthesis Approach for Automatic Vertebral Shape Identification in Clinical QCT. In: Brox T., Bruhn A., Fritz M. (eds) Pattern Recognition. GCPR 2018. Lecture Notes in Computer Science, vol 11269. Springer, Cham" (preprint here).

    Model Creation Step Example output
    Model creation GUI Example cortex identification result

     

    The library is written in C++ but also has a C and Matlab bindings.

    Binary releases are available for Linux, Windows 10 and macOS.

    The source code can be found in this GitHub repository

    Contact/Maintainer: Stefan Reinhold


  • Simulation of Plenoptic Cameras


    In order to create ground truth data for algorithms working on plenoptic camera data, we built a model of a reconfigurable plenoptic camera for Blender. Contrary to other simulations, our model does not heavily simplify the optical system, but contains an objective as well as a microlens array and therefore shows correct image degradation effects.

    Plenoptic Simulation

    The .blend file and configuration instructions are available in this GitHub repository and the corresponding Wiki.

    Contact/Maintainer: Arne Petersen, Tim Michels


  • Ground Truth Creation for Camera Calibration


    In order to evaluate calibration methods for plenoptic or conventional cameras it can be useful to have realistic synthetic ground truth data, i.e. realistic renderings of calibration patterns and the ground truth pixel positions of the pattern's points of interest as well as the corresponding 3D position. To this end we extended our camera simulation in Blender and built a corner extraction tool.

    Screenshot

    The tool as well as instructions on the usage can be found in this GitLab repository.

    Contact/Maintainer: Tim Michels


  • Multi Camera Calibration


    Since we often face the problem of calibrating setups containing multiple cameras with different properties, we built an easy-to-use calibration tool based on the detection of ChArUco boards, a combination of traditional checkerboards and ArUco markers. These markers allow the calibration boards to be partly occluded and accordingly the pattern can also be recognized if it is only partly located within a camera's field of view.

    Calibration Tool

    The tool as well as instructions on the usage can be found in this GitLab repository and its Wiki.

    Contact/Maintainer: Tim Michels


  • Kinect/Projector Calibration


    Several augmented reality applications like our ARSandbox require a system consisting of a depth camera and a projector to be calibrated. For this purpose we build a small tool facilitating the whole procedure.

    Screenshot

    The tool as well as instructions on the usage can be found in this GitLab repository.

    Contact/Maintainer: Tim Michels


  • Batch File Renamer


    When dealing with a large number of files, e.g. images captured with our light field rig, the renaming process for adjusting the filenames according to the requirements of some MatLab or Github toolbox can be time-consuming. To simplify this process, we built a small tool which uses a tag/delimiter system to analyze and rename files.

    Renamer Tool

    The tool as well as instructions on the usage can be found in this GitLab repository.

    Contact/Maintainer: Tim Michels