Paper für die International Conference on 3D Vision 2020 akzeptiert

Das Paper mit dem Titel  „Error Bounds of Projection Models in Weakly Supervised 3D Human Pose Estimation“ von Nikolas Klug, Moritz Einfalt, Stephan Brehm und Rainer Lienhart wurde für die International Conference on 3D Vision (3DV) 2020 akzeptiert.

In diesem Paper untersuchen die Autoren quantitativ, welche negativen Auswirkungen vereinfachte Abbildungsmodelle auf die Schätzung menschlicher 3D Pose in Bildern haben. Das Paper wird im Rahmen der Online-Konferenz vom 25. -     27.11.2020 vorgestellt.



The current state-of-the-art in monocular 3D human pose estimation is heavily influenced by weakly supervised methods. These allow 2D labels to be used to learn effective 3D human pose recovery either directly from images or via 2D-to-3D pose uplifting. In this paper we present a detailed analysis of the most commonly used simplified projection models, which relate the estimated 3D pose representation to 2D labels: normalized perspective and weak perspective projections. Specifically, we derive theoretical lower bound errors for those projection models under the commonly used mean per-joint position error (MPJPE). Additionally, we show how the normalized perspective projection can be replaced to avoid this guaranteed minimal error. We evaluate the derived lower bounds on the most commonly used 3D human pose estimation benchmark datasets. Our results show that both projection models lead to an inherent minimal error between $19.3$mm and $54.7$mm, even after alignment in position and scale. This is a considerable share when comparing with recent state-of-the-art results. Our paper thus establishes a theoretical baseline that shows the importance of suitable projection models in weakly supervised 3D human pose estimation.


Nikolas Klug, Moritz Einfalt, Stephan Brehm, Rainer Lienhart.
Error Bounds of Projection Models in Weakly Supervised 3D Human Pose Estimation.
2020 International Conference on 3D Vision (3DV), IEEE. Fukuoka, Japan, November 2020. To appear. [ arXiv]


Beispiele für menschliche 3D Posen (rot) aus Human3.6M, MPI-INF-3DHP und CMU Panoptic sowie die bestmöglich rekonstruierten 3D Pose unter einer normalisiert-perspektivischen (blau) und einer schwach-perspektivischen Projektion (grün) © Universität Augsburg