Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Revisiting Rotation Averaging : Uncertainties and Robust Losses

Zhang, Ganlin ; Larsson, Viktor LU and Barath, Daniel (2023) 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023 In Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition 2023-June. p.17215-17224
Abstract

In this paper, we revisit the rotation averaging problem applied in global Structure-from-Motion pipelines. We argue that the main problem of current methods is the minimized cost function that is only weakly connected with the input data via the estimated epipolar geometries. We propose to better model the underlying noise distributions by directly propagating the uncertainty from the point correspondences into the rotation averaging. Such uncertainties are obtained for free by considering the Jacobians of two-view refinements. Moreover, we explore integrating a variant of the MAGSAC loss into the rotation averaging problem, instead of using classical robust losses employed in current frameworks. The proposed method leads to results... (More)

In this paper, we revisit the rotation averaging problem applied in global Structure-from-Motion pipelines. We argue that the main problem of current methods is the minimized cost function that is only weakly connected with the input data via the estimated epipolar geometries. We propose to better model the underlying noise distributions by directly propagating the uncertainty from the point correspondences into the rotation averaging. Such uncertainties are obtained for free by considering the Jacobians of two-view refinements. Moreover, we explore integrating a variant of the MAGSAC loss into the rotation averaging problem, instead of using classical robust losses employed in current frameworks. The proposed method leads to results superior to baselines, in terms of accuracy, on large-scale public benchmarks. The code is public. https://github.com/zhangganlin/GlobalSfMpy

(Less)
Please use this url to cite or link to this publication:
author
; and
organization
publishing date
type
Chapter in Book/Report/Conference proceeding
publication status
published
subject
keywords
3D from multi-view and sensors
host publication
Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023
series title
Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
volume
2023-June
pages
10 pages
publisher
IEEE Computer Society
conference name
2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023
conference location
Vancouver, Canada
conference dates
2023-06-18 - 2023-06-22
external identifiers
  • scopus:85173945874
ISSN
1063-6919
ISBN
9798350301298
DOI
10.1109/CVPR52729.2023.01651
language
English
LU publication?
yes
id
cb64a3a3-b4d2-4caa-b1c6-3472dce56bbf
date added to LUP
2023-12-15 09:57:17
date last changed
2023-12-15 09:59:12
@inproceedings{cb64a3a3-b4d2-4caa-b1c6-3472dce56bbf,
  abstract     = {{<p>In this paper, we revisit the rotation averaging problem applied in global Structure-from-Motion pipelines. We argue that the main problem of current methods is the minimized cost function that is only weakly connected with the input data via the estimated epipolar geometries. We propose to better model the underlying noise distributions by directly propagating the uncertainty from the point correspondences into the rotation averaging. Such uncertainties are obtained for free by considering the Jacobians of two-view refinements. Moreover, we explore integrating a variant of the MAGSAC loss into the rotation averaging problem, instead of using classical robust losses employed in current frameworks. The proposed method leads to results superior to baselines, in terms of accuracy, on large-scale public benchmarks. The code is public. https://github.com/zhangganlin/GlobalSfMpy</p>}},
  author       = {{Zhang, Ganlin and Larsson, Viktor and Barath, Daniel}},
  booktitle    = {{Proceedings - 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2023}},
  isbn         = {{9798350301298}},
  issn         = {{1063-6919}},
  keywords     = {{3D from multi-view and sensors}},
  language     = {{eng}},
  pages        = {{17215--17224}},
  publisher    = {{IEEE Computer Society}},
  series       = {{Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition}},
  title        = {{Revisiting Rotation Averaging : Uncertainties and Robust Losses}},
  url          = {{http://dx.doi.org/10.1109/CVPR52729.2023.01651}},
  doi          = {{10.1109/CVPR52729.2023.01651}},
  volume       = {{2023-June}},
  year         = {{2023}},
}