Feature

“Hm. What kind of company is that, Fraunhofer IGD”, was my first thought when I was told that they had a job opening in research coaching & management. As a “die-hard” academic scientist, I wasn’t that keen on “shifting from academics to industry”. Well, actually that didn’t happen, although I started at Fraunhofer IGD last year.

Fraunhofer

The first reaction of most people in my environment is “hey - aren’t those the people who invented the MP3 format?” Yes, indeed, they did – well, it was at another institute, but nevertheless, when you buy your MP3s I now also benefit from the royalties! The success story of MP3 is a template of the way of working of the Fraunhofer Gesellschaft (group).

In 1814, Joseph von Fraunhofer invented the spectroscope, and discovered 574 dark lines appearing in the solar spectrum. These were later shown to be atomic absorption lines. These lines are still called Fraunhofer lines in his honour (see also en.wikipedia.org/wiki/Joseph_von_Fraunhofer). Later he became director of the Optical Institute at Benediktbeuern, a secularised Benedictine monastery devoted to glass making, where he had discovered how to make the world's finest optical glass and invented incredibly precise methods for measuring dispersion. In his spirit, the close to 60 Fraunhofer institutes perform applied scientific research building bridges between industry and academia.

Darmstadt

One of the institutes, Fraunhofer IGD, is located in Darmstadt, a bit south of Frankfurt with its huge international airport (for those familiar with international flights, but lacking some detailed geography of Germany!). Together with several groups at the Technical University of Darmstadt, Fraunhofer IGD forms a centre of Visual Computing, working on all aspects of computer vision, learning, graphics, and pattern recognition. They publish their work in the major journals and present it at important conferences – a reason for me to move here as I am also affiliated to the Technical University of Darmstadt! One of the groups is headed by the director of Fraunhofer IGD, guaranteeing a steady flow of state-of-the-art methodologies and relevant applications in both directions – a standard Fraunhofer construction, by the way.

Not only do we visit conferences, we also organise them. For instance, September 22-24, 2010, the 32nd annual pattern recognition symposium of the German Association for Pattern Recognition, the German section of the IAPR, will take place in Darmstadt – see www.dagm2010.org/ for more details. It would be a great opportunity to see what is going on here! If those dates don’t fit, you can visit us October 15-17, 2010, during the 6th International Conference on Intelligent Information Hiding and Multimedia Signal Processing IIHMSP 2010, bit.kuas.edu.tw/~iihmsp10/.

TUD

If you don’t want to wait until 2010, let me first give a brief overview of some recent work of the TU groups at Computer Science that work on Visual Computing. Their research focuses on the detection of people / pedestrians / objects and recognition of activities. Furthermore, we have parallel computing, investigating multi-view stereo for large image collections like Flickr, and statistical structure learning. Besides these topics, we have groups working on visual analytics and medical computing. To find out more, just Google for the names Schiele, Schindler, Roth, or Goesele in combination with Darmstadt (and feel free to Google me as well!).

Fraunhofer IGD: Vision

As is already clear from the previous section, a large spectrum of visual computing is covered in Darmstadt. As IGD stands for “Institut für Graphische Datenverarbeitung” – Institute for Computer Graphics Research, you can imagine that Computer Graphics is an important research topic – also at the TU. You’re right. That’s how it started, but nowadays the scope of the institute has become much broader.

Traditionally, Computer Graphics, with topics like virtual and augmented reality, rendering, and visualization, has been separated from Computer Vision, dealing with pattern recognition, image segmentation, and image understanding. One can think of the former as using a model and the latter as trying to find one. Nowadays, almost all the current challenges in research and industry appear at the interface and in the interdependency of graphics and vision. It is therefore one of the three research lines of Fraunhofer IGD: Confluence of Graphics and Vision.

Another research line, Semantics in the Modeling Process, extends the classical level-of-detail approach in modeling to the inclusion of a semantic level-of-detail. For instance, when a car is modeled, it should be clear at all description levels (e.g. wheels) that we are dealing with a car and not a motorbike. Therefore, a new approach to modeling including a semantic level-of-detail besides the classical level-of-detail is required.

The third line is Generalized Digital Documents. In the context of classical documents, we address questions which cannot even be worded clearly in the context of multimedia data. For instance, a thumbnail might be agreed upon as the abstract of an image. But the definition of the abstract of a complex 3D model or a video is not clear in an application independent sense.

Research areas

These three research lines are found in the individual departments in Darmstadt that focus on interactive multimedia appliances, industrial applications, real-time solutions for simulation and visual analytics, virtual and augmented reality, graphic information systems, 3D knowledge worlds and semantics visualization, cognitive computing & medical imaging, and security technology.

Next to those there are Fraunhofer IGD branches in Rostock, Singapore, and Graz (Austria) with interests in scientific and medical visualization, environments for life sciences, virtual and augmented environments for medical applications, new media for education and cultural heritage, 3D modeling and reconstruction of incident scenes, teaching and learning with the internet and new media, perceptive and affective applications which assist and support people in their job or leisure time, and maritime graphics.

Outlook

So what does this nice broad list of research areas tell us? Well, for instance that a lot of interesting research is going on, inspired by developing computing possibilities. Think big! Traditional pattern recognition areas like irises and fingerprints are of importance to security technology, e.g. an integrating person recognition system that may partially replace passport checks at airports. Or one would like to find certain (a priori undefined) patterns in huge (visualized) data sets. Medical computing still needs appropriate algorithms. Cultural heritage is booming. The amount of sensor-produced data is exploding – we need methods to analyze, visualize, and store them properly.

There’s enough work to be done!