A novel metric for 6D pose estimation
-
Tobias Niedermaier
M.Sc. Tobias Niedermaier is a doctoral candidate in the field of sensor fusion at the Institute for Automation and Applied Computer Science at the Karlsruhe Institute of Technology, and the Institute for Artificial Intelligence at the Ravensburg-Weingarten University of Applied Sciences. Research Interests: Sensor fusion, machine learning, object detection, deep learning.
,
Felix Berens
M.Sc. Felix Berens is a doctoral candidate in the field of sensor fusion for autonomous driving at the Institue for Automation and Applied Computer Science at the Karlsruhe Institute of Technology, and the Institute for Artificial Intelligence at the Ravensburg-Weingarten University of Applied Sciences. Research Interests: Sensor fusion, sensor placement, machine learning, object detection.
,
Markus Reischl
Prof. Dr.-Ing. Markus Reischl is head of the research group ''Machine Learning for High-Through put and Mechatronics'' of the Institute for Automation and Applied Computer Science at the Karlsruhe Institute of Technology. Research Interests: Man-machine
Prof. Dr. rer. nat. Stefan Elser works as a professor for autonomous driving at the Ravensburg-Weingarten University of Applied Sciences. Research Interests: Machine learning, object detection, sensor fusion and their applications in autonomous driving.
Abstract
Current state-of-the-art evaluation methods for 6D pose estimation have several significant drawbacks. Existing error metrics can produce near-zero errors for poor pose estimations and are heavily dependent on the object point cloud used, resulting in vastly different outcomes for different objects. Furthermore, false detections are not considered at all. Evaluating pose estimators is crucial as it directly impacts the reliability and effectiveness of applications in robotics, augmented reality, and object manipulation tasks. Accurate evaluation ensures that pose estimators can be trusted to perform well in real-world scenarios, leading to better system performance and user satisfaction. In this paper, we conduct experiments to provide insights into how these metrics behave under isolated errors. We also introduce a novel evaluation approach with a metric independent of point clouds, making it applicable to a broader range of use cases than current metrics.
Zusammenfassung
Aktuelle Methoden zur Bewertung der 6D-Posenschätzung weisen mehrere erhebliche Nachteile auf. Bestehende Fehlerkennzahlen können bei schlechten Posenabschätzungen nahezu fehlerfreie Ergebnisse liefern und sind stark von der verwendeten Objektpunktwolke abhängig, was zu stark unterschiedlichen Ergebnissen für verschiedene Objekte führt. Darüber hinaus werden falsche Erkennungen überhaupt nicht berücksichtigt. Die Bewertung von Posenabschätzern ist entscheidend, da sie die Zuverlässigkeit und Effektivität von Anwendungen in der Robotik, Augmented Reality und Objektmanipulation direkt beeinflusst. Eine genaue Bewertung stellt sicher, dass Posenabschätzer in realen Szenarien zuverlässig funktionieren, was zu einer besseren Systemleistung und Benutzerzufriedenheit führt. In dieser Arbeit führen wir Experimente durch, um Einblicke in das Verhalten dieser Kennzahlen bei isolierten Fehlern zu geben. Wir stellen auch einen neuartigen Bewertungsansatz mit einer Kennzahl vor, die unabhängig von Punktwolken ist und somit auf ein breiteres Spektrum von Anwendungsfällen anwendbar ist als aktuelle Kennzahlen.
About the authors
M.Sc. Tobias Niedermaier is a doctoral candidate in the field of sensor fusion at the Institute for Automation and Applied Computer Science at the Karlsruhe Institute of Technology, and the Institute for Artificial Intelligence at the Ravensburg-Weingarten University of Applied Sciences. Research Interests: Sensor fusion, machine learning, object detection, deep learning.
M.Sc. Felix Berens is a doctoral candidate in the field of sensor fusion for autonomous driving at the Institue for Automation and Applied Computer Science at the Karlsruhe Institute of Technology, and the Institute for Artificial Intelligence at the Ravensburg-Weingarten University of Applied Sciences. Research Interests: Sensor fusion, sensor placement, machine learning, object detection.
Prof. Dr.-Ing. Markus Reischl is head of the research group ''Machine Learning for High-Through put and Mechatronics'' of the Institute for Automation and Applied Computer Science at the Karlsruhe Institute of Technology. Research Interests: Man-machine
Prof. Dr. rer. nat. Stefan Elser works as a professor for autonomous driving at the Ravensburg-Weingarten University of Applied Sciences. Research Interests: Machine learning, object detection, sensor fusion and their applications in autonomous driving.
-
Research ethics: Not applicable.
-
Informed consent: Not applicable.
-
Author contributions: All authors have accepted responsibility for the entire content of this manuscript and approved its submission.
-
Use of Large Language Models, AI and Machine Learning Tools: None declared.
-
Conflict of interest: The authors state no conflict of interest.
-
Research funding: This research was partially funded by Zentrale Innovationsprogramm Mittelstand (ZIM). Grant number: 16KN096031.
-
Data availability: Not applicable.
References
[1] A. Blank, et al.., “6DoF pose-estimation pipeline for texture-less industrial components in bin picking applications,” in 2019 European Conference on Mobile Robots (ECMR), IEEE, 2019, pp. 1–7.10.1109/ECMR.2019.8870920Search in Google Scholar
[2] A. Cordeiro, et al.., “Bin picking for ship-building logistics using perception and grasping systems,” Robotics, vol. 12, no. 1, p. 15, 2023. https://doi.org/10.3390/robotics12010015.Search in Google Scholar
[3] J. Li, et al.., “ERegPose: an explicit regression based 6D pose estimation for snake-like wrist-type surgical instruments,” Int. J. Med. Robot. Comput. Assist. Surg., vol. 20, no. 3, p. e2640, 2024. https://doi.org/10.1002/rcs.2640.Search in Google Scholar PubMed
[4] Y. Xiang, et al.., “PoseCNN: a convolutional neural network for 6D object pose estimation in cluttered scenes,” CoRR abs/1711.00199, 2017, arXiv: 1711.00199.10.15607/RSS.2018.XIV.019Search in Google Scholar
[5] S. Hinterstoisser, et al.., “Model based training, detection and pose estimation of texture-less 3D objects in heavily cluttered scenes,” in Computer Vision – ACCV 2012, K. M. Lee, et al.., Eds., Berlin, Heidelberg, Springer Berlin Heidelberg, 2013, pp. 548–562.10.1007/978-3-642-37331-2_42Search in Google Scholar
[6] Y. He, et al.., “FFB6D: a full flow bidirectional fusion network for 6D pose estimation,” in 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021, pp. 3002–3012.10.1109/CVPR46437.2021.00302Search in Google Scholar
[7] M. Stoiber, et al.., “Fusing visual appearance and geometry for multi-modality 6DoF object tracking,” 2023, arXiv: 2302.11458.10.1109/IROS55552.2023.10341961Search in Google Scholar
[8] G. Marullo, et al.., “6D object position estimation from 2D images: a literature review,” Multimed. Tool. Appl., vol. 82, no. 16, pp. 24605–24643, 2023. https://doi.org/10.1007/s11042-022-14213-z.Search in Google Scholar
[9] J. Guan, et al.., “A survey of 6DoF object pose estimation methods for different application scenarios,” Sensors, vol. 24, no. 4, p. 1076, 2024. https://doi.org/10.3390/s24041076.Search in Google Scholar PubMed PubMed Central
[10] C. Wang, et al.., “Densefusion: 6d object pose estimation by iterative dense fusion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 3343–3352.10.1109/CVPR.2019.00346Search in Google Scholar
[11] B. Drost, et al.., “Model globally, match locally: efficient and robust 3D object recognition,” in 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, IEEE, 2010, pp. 998–1005.10.1109/CVPR.2010.5540108Search in Google Scholar
[12] T. Hodaň, J. Matas, and Š. Obdržálek, “On evaluation of 6D object pose estimation,” in Computer Vision–ECCV 2016 Workshops: Amsterdam, The Netherlands, October 8–10 and 15–16, 2016, Proceedings, Part III 14, Springer, 2016, pp. 606–619.10.1007/978-3-319-49409-8_52Search in Google Scholar
[13] H. Yu, et al.., “Robust robot pose estimation for challenging scenes with an RGB-D camera,” IEEE Sens. J., vol. 19, no. 6, pp. 2217–2229, 2019. https://doi.org/10.1109/JSEN.2018.2884321.Search in Google Scholar
[14] E. Muñoz, et al.., “Fast 6D pose from a single RGB image using cascaded forests templates,” in 2016 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2016, pp. 4062–4069.10.1109/IROS.2016.7759598Search in Google Scholar
[15] D. Q. Huynh, “Metrics for 3D rotations: comparison and analysis,” J. Math. Imag. Vis., vol. 35, no. 2, pp. 155–164, 2009. https://doi.org/10.1007/s10851-009-0161-2.Search in Google Scholar
[16] A. Palazzolo, “Formalism for the rotation matrix of rotations about an arbitrary axis,” Am. J. Phys., vol. 44, no. 1, pp. 63–67, 1976. https://doi.org/10.1119/1.10140.Search in Google Scholar
[17] A. Morawiec, Orientations and Rotations: Computations in Crystallographic Textures, Heidelberg, Germany, Springer Berlin, 2013.Search in Google Scholar
[18] M. Kazhdan, T. Funkhouser, and S. Rusinkiewicz, “Symmetry descriptors and 3D shape matching,” in Proceedings of the 2004 Eurographics/ACM SIGGRAPH Symposium on Geometry Processing, 2004, pp. 115–123.10.1145/1057432.1057448Search in Google Scholar
[19] E. Corona, K. Kundu, and S. Fidler, “Pose estimation for objects with rotational symmetry,” CoRR abs/1810.05780, 2018, arXiv: 1810.05780, https://doi.org/10.1109/iros.2018.8594282.Search in Google Scholar
© 2024 Walter de Gruyter GmbH, Berlin/Boston
Articles in the same Issue
- Frontmatter
- Methods
- Quantifying and combining uncertainty for improving the behavior of Digital Twin Systems
- Dynamic reconfiguration of application and communication in Industry 4.0 with Digital Twin for enhancing flexibility
- Modelling of the NAMUR NE 178 “Verification of Request” concept using Pro-Active Asset Administration Shells
- A novel metric for 6D pose estimation
- Anwendungen
- Coordinated operation of pumped hydro energy storage with reversible pump turbine and co-located battery energy storage system
- Tools
- Test engines for the asset administration shell
Articles in the same Issue
- Frontmatter
- Methods
- Quantifying and combining uncertainty for improving the behavior of Digital Twin Systems
- Dynamic reconfiguration of application and communication in Industry 4.0 with Digital Twin for enhancing flexibility
- Modelling of the NAMUR NE 178 “Verification of Request” concept using Pro-Active Asset Administration Shells
- A novel metric for 6D pose estimation
- Anwendungen
- Coordinated operation of pumped hydro energy storage with reversible pump turbine and co-located battery energy storage system
- Tools
- Test engines for the asset administration shell
