This project aims at developing novel algorithms for optical flow (2D motion) and scene flow (3D motion) estimation. Instead of relying on fixed assumptions, the goal is to adaptively integrate prior knowledge and other available information in order to design approaches that are not only highly accurate but also generalize well across datasets.
Fig. 1: Optical flow estimation on the automotive KITTI dataset using the method of Maurer et al. (BMVC 2018).
How can we develop algorithms for adaptive motion estimation of high accuracy and generalizability that are applicable in the wild?
To which extent can concepts be generalized from the 2D to the 3D domain and how can additional constraints imposed by the 3D world improve the estimation?
How can we transfer the benefits of adaptive algorithms to specific applications?
FOR SCIENTISTS
Projects
People
Publications
Graduate School
Equal Opportunity
FOR PUPILS
PRESS AND MEDIA