ML Anthology

A Case for Using Rotation Invariant Features in State of the Art Feature Matchers

Georg Bökman, Fredrik Kahl
CVPRW 2022 pp. 5106-5115
doi:10.1109/CVPRW56347.2022.00559 /cvprw/2022/bokman2022cvprw-case/

Abstract

The aim of this paper is to demonstrate that a state of the art feature matcher (LoFTR) can be made more robust to rotations by simply replacing the backbone CNN with a steerable CNN which is equivariant to translations and image rotations. It is experimentally shown that this boost is obtained without reducing performance on ordinary illumination and viewpoint matching sequences.

PDF CVPRW Semantic Scholar

Cite

Text

Bökman and Kahl. "A Case for Using Rotation Invariant Features in State of the Art Feature Matchers." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2022. doi:10.1109/CVPRW56347.2022.00559

Markdown

[Bökman and Kahl. "A Case for Using Rotation Invariant Features in State of the Art Feature Matchers." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2022.](https://mlanthology.org/cvprw/2022/bokman2022cvprw-case/) doi:10.1109/CVPRW56347.2022.00559

BibTeX

@inproceedings{bokman2022cvprw-case,
  title     = {{A Case for Using Rotation Invariant Features in State of the Art Feature Matchers}},
  author    = {Bökman, Georg and Kahl, Fredrik},
  booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
  year      = {2022},
  pages     = {5106-5115},
  doi       = {10.1109/CVPRW56347.2022.00559},
  url       = {https://mlanthology.org/cvprw/2022/bokman2022cvprw-case/}
}