Advertisement

Well Convergent and Computationally Efficient Quaternion Loss

Conference paper
  • 370 Downloads
Part of the Advances in Intelligent Systems and Computing book series (AISC, volume 1196)

Abstract

Rotation estimation, i.e. the ability to predict angles describing 3D positioned object, is an omnipresent problem in computer vision, computer graphics and 3D object detection task in automotive industry. Deep learning algorithms usually parameterise rotation using only the yaw angle. This paper presents detailed comparison of several 3D rotation distance functions using quaternions. We propose a computationally efficient quaternion loss function for neural network training. We conclude that function respects the topology of SO(3) and is bi-invariant. We also show the geometrical representation of presented functions. Lastly, we evaluate the effectiveness of the proposed loss function and compare its performance with other methods using a public large-scale dataset.

Keywords

Optimization Quaternions Sensors Neural networks Automotive 

Notes

Acknowledgment

Research was funded by Polish Ministry of Science and Higher Education Project No. 0014/DW/2018/02 and carried out in cooperation of Aptiv Services Poland S.A. - Technical Center Kraków and AGH University of Science and Technology - Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering.

References

  1. 1.
    Altmann, S.: Rotations, Quaternions, and Double Groups. Dover Publications (1986)Google Scholar
  2. 2.
    Angulo, J.: Riemannian l (p) averaging on lie group of nonzero quaternions. Adv. Appl. Clifford Algebras 24 (2014).  https://doi.org/10.1007/s00006-013-0432-2
  3. 3.
    Arkhangel’skii, A.V., Pontryagin, L.S.: General Topology I. Springer, Heidelberg (1990)CrossRefGoogle Scholar
  4. 4.
    Caesar, H., Bankiti, V., Lang, A.H., Vora, S., Liong, V.E., Xu, Q., Krishnan, A., Pan, Y., Baldan, G., Beijbom, O.: nuScenes: a multimodal dataset for autonomous driving (2019)Google Scholar
  5. 5.
    Dam, E.B., Koch, M., Lillholm, M.: Quaternions, interpolation and animation. Tech. rep. 1, Department of Computer Science University of Copenhagen, Denmark (1998)Google Scholar
  6. 6.
    Euler, L.: Decouverte d’un nouveau principe de mecanique. In: Opera Omnia 2, vol. 5, pp. 81–108. Orell Füssli (1753)Google Scholar
  7. 7.
    Geiger, A., Lenz, P., Stiller, C., Urtasun, R.: Vision meets robotics: the kitti dataset. I. J. Robotic Res. 32(11), 1231–1237 (2013). http://dblp.uni-trier.de/db/journals/ijrr/ijrr32.html#GeigerLSU13
  8. 8.
    Hamilton, W.R.: Lectures on quaternions. Hodges and Smith, Dublin (1853)Google Scholar
  9. 9.
    Huynh, D.Q.: Metrics for 3D rotations: comparison and analysis. J. Math. Imaging Vis. 35(2), 155–164 (2009)MathSciNetCrossRefGoogle Scholar
  10. 10.
    Kuffner, J.: Effective sampling and distance metrics for 3D rigid body path planning. In: Proceedings - IEEE International Conference on Robotics and Automation, vol. 4, pp. 3993–3998 (2004).  https://doi.org/10.1109/ROBOT.2004.1308895
  11. 11.
    Ravani, B., Roth, B.: Smooth invariant interpolation of rotations. J. Mech. Transm. Autom. Des. 105(3), 460–467 (1983)CrossRefGoogle Scholar
  12. 12.
    Redmon, J., Divvala, S., Girshick, R., Farhadi, A.: You only look once: unified, real-time object detection. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 779–788 (2016)Google Scholar
  13. 13.
    Redmon, J., Farhadi, A.: Yolo9000: better, faster, stronger. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 7263–7271 (2017)Google Scholar
  14. 14.
    Redmon, J., Farhadi, A.: Yolov3: an incremental improvement. arXiv preprint arXiv:1804.02767 (2018)
  15. 15.
    Rossmann, W.: Lie Groups: An Introduction Through Linear Groups. No. 9780199202515 in OUP Catalogue. Oxford University Press (2006). https://ideas.repec.org/b/oxp/obooks/9780199202515.html
  16. 16.
    Wunsch, P., Winkler, S., Hirzinger, G.: Real-time pose estimation of 3D objects from camera images using neural networks. In: Proceedings of International Conference on Robotics and Automation, vol. 4, pp. 3232–3237 (1997).  https://doi.org/10.1109/ROBOT.1997.606781

Copyright information

© Springer Nature Switzerland AG 2020

Authors and Affiliations

  1. 1.Department of Automatic Control and RoboticsAGH University of Science and TechnologyKrakówPoland
  2. 2.APTIV Services Poland S.A.KrakówPoland

Personalised recommendations