Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Single Shot 6D Object Pose Estimation

: Kleeberger, Kilian; Huber, Marco

Fulltext ()

Institute of Electrical and Electronics Engineers -IEEE-; IEEE Robotics and Automation Society:
IEEE International Conference on Robotics and Automation, ICRA 2020 : 31 May - 31 August 2020, Virtuell, Paris, France
Piscataway, NJ: IEEE, 2020
ISBN: 978-1-7281-7395-5
ISBN: 978-1-7281-7394-8
ISBN: 978-1-7281-7396-2
International Conference on Robotics and Automation (ICRA) <2020, Online>
Conference Paper, Electronic Publication
Fraunhofer IPA ()
Bin-Picking; Deep Learning; Künstliche Intelligenz; Maschinelles Lernen; Robotik

In this paper, we introduce a novel single shot approach for 6D object pose estimation of rigid objects based on depth images. For this purpose, a fully convolutional neural network is employed, where the 3D input data is spatially discretized and pose estimation is considered as a regression task that is solved locally on the resulting volume elements. With 65 fps on a GPU, our Object Pose Network (OP-Net) is extremely fast, is optimized end-to-end, and estimates the 6D pose of multiple objects in the image simultaneously. Our approach does not require manually 6D pose-annotated real-world datasets and transfers to the real world, although being entirely trained on synthetic data. The proposed method is evaluated on public benchmark datasets, where we can demonstrate that state-of-the-art methods are significantly outperformed.