Computer Science Takes Second Place in Stereo Vision Competition at KIT Germany

24 September 2012

Stereo vision is a significant component of today's computer vision solutions, as used in advanced applications for robotics, industrial automation, generation of three-dimensional city maps or models of buildings, or in vision-based driver assistance. Stereo vision algorithms have been designed for nearly 40 years in hundreds of research groups worldwide. For comparing the performance of those algorithms, benchmark websites offer data sets which can be used as an input by researchers, and their output is then compared to the known true values, unknown to the researchers, and only known to the operators of the website.

The Karlsruhe Institute of Technology, Germany, and the Toyota Technological Institute at Chicago, USA, provide such a benchmark website, thus defining a competition for researchers in computer vision worldwide. For more information, see the KITTI Vision Benchmark Suite website.

The provided test data are especially challenging, because they are recorded in a car in the real world under different lighting and weather conditions. The difficulty of the data also means that not many researchers actually achieve the level of accuracy in their stereo reconstructions to be able to compete reasonably with already submitted results.

On 21 September 2012, the ranking of stereo vision algorithms on this website is as follows:

  1. Researchers from the Toyota Technological Institute at Chicago, USA
  2. Researchers from The University of Auckland.
  3. A team of researchers from the Daimler Research Centre, Germany, and
    Graz University.
  4. Anonymous.
  5. A researcher from the German Air and Space Institute.

Major players in the field of stereo vision compete here for the first places for best performing stereo vision algorithms on challenging real-world data, typically supported by well-funded research programs.
It is a major achievement that researchers of The University of Auckland joined those leading teams, being currently ranked second.

The design and implementation of The University of Auckland's successful stereo analysis program was done by Simon Hermann, PhD student in the Computer Science Department, Tamaki Innovation Campus, supervised by Professor Reinhard Klette. Simon and Reinhard will present their stereo vision solution at the Asian Computer Vision Conference this November in Korea. It is also a great achievement to have an accepted presentation at this top-ranked international computer vision conference.

For more information about this research, please contact:
Professor Reinhard Klette