Creating ground truth for optical flow in natural outdoor environments seems almost impossible.
In this talk, I will propose two approaches we are currently investigating.
The first approach is to use semi-automatic vision algorithms as is done for example in movie postproduction to create “weak” ground truth.
The second approach is to evaluate the properties of today’s computer graphic rendering systems with respect to their ability to generate images close to the real world.
Finally, I will discuss the problem of defining performance measures and benchmarking with respect to correspondence estimation and related algorithms.
The datasets concerning this research line will be made available at:
Daniel Kondermann studied Computer Sciences and received his Diploma degree in 2006 from the TU Ilmenau.
He obtained a Doctoral degree in 2009 for his work on “Modular Optical Flow Estimation with Applications to Fluid Dynamics”
at the Interdisciplinary Center of Scientific Computing (IWR) of the University of Heidelberg.
Since then, he has been Postdoctoral fellow at the Heidelberg Collaboratory for Image Processing (HCI),
an Industry on Campus Institution of Heidelberg University with the participation of several companies.
His research interests include image processing, computer vision and artificial intelligence research in general as well as optical flow, stereo vision, depth imaging, 3d reconstruction and performance evaluation in particular.