Showing 1–1 of 1 results for author: Leiss, K

Automated Scene Flow Data Generation for Training and Verification

Authors: Oliver Wasenmüller, René Schuster, Didier Stricker, Karl Leiss, Jürger Pfister, Oleksandra Ganus, Julian Tatsch, Artem Savkin, Nikolas Brasch

Abstract: Scene flow describes the 3D position as well as the 3D motion of each pixel in an image. Such algorithms are the basis for many state-of-the-art autonomous or automated driving functions. For verification and training large amounts of ground truth data is required, which is not available for real data. In this paper, we demonstrate a technology to create synthetic data with dense and precise scene… ▽ More Scene flow describes the 3D position as well as the 3D motion of each pixel in an image. Such algorithms are the basis for many state-of-the-art autonomous or automated driving functions. For verification and training large amounts of ground truth data is required, which is not available for real data. In this paper, we demonstrate a technology to create synthetic data with dense and precise scene flow ground truth. △ Less

Submitted 31 August, 2018; v1 submitted 30 August, 2018; originally announced August 2018.