2016/12/31 by Benjamin Ummenhofer, Huizhong Zhou, Jonas Uhrig +4 · 2 citations
Computer Science · #cs.CV
paper · pdf · doi:10.1109/cvpr.2017.596
Camera ready version for CVPR 2017. Supplementary material included. Project page: http://lmb.informatik.uni-freiburg.de/people/ummenhof/depthmotionnet/
arxiv created 2017/04/11 · arxiv updated 2018/01/18
In this paper we formulate structure from motion as a learning problem. We train a convolutional network end-to-end to compute depth and camera motion from successive, unconstrained image pairs. The architecture is composed of multiple stacked encoder-decoder networks, the core part being an iterative network that is able to improve its own predictions. The network estimates not only depth and motion, but additionally surface normals, optical flow between the images and confidence of the matching. A crucial component of the approach is a training loss based on spatial relative differences. Compared to traditional two-frame structure from motion methods, results are more accurate and more robust. In contrast to the popular depth-from-single-image networks, DeMoN learns the concept of matching and, thus, better generalizes to structures not seen during training.