The research group “Machine Learning for High-Throughput Methods and Mechatronics” (ML4HOME) deals with the analysis of data being generated automatically by mechatronic systems in big amounts. Examples are automated microscopes, X-ray diffractometers and autonomous vehicles. Our focus is the generalization of big data-sets consisting of a lot of data points (high throughput), whose extent requires mandatory automatic processing.

We aim for modelling processes to identify relations and outliers in big data sets (images, time series etc.) to automatically quantify experiments, predict system conditions or give instant feedback about inconsistent measurements. Visualization routines are commonly applied to clarify the results.

To ensure the usability of all developed algorithms and to enable external researchers to process their data, we integrate our methods typically into stand-alone software tools. Modelling is mostly done using Python’s libraries scikit-learn, PyTorch and TensorFlow, easy processing is done using MATLAB.

Helmholtz Program (PoF IV): Natural, Artificial and Cognitive Information Processing (NACIP)
(BioInterfaces in Technology and Medicine (BIFTM), PoF III)

Projects:

• Automated Image and Data Analysis
• Bildbasierte Fahrbahnbewertung

Completed Projects:

• TELMYOS
• Establishment of a new method to interactively segment and visualize complex 3D image data