SCENES: Subpixel Correspondence Estimation With Epipolar Supervision

Dominik A. Kloepfer*, Joao F. Henriques, Dylan Campbell

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Extracting point correspondences from two or more views of a scene is a fundamental computer vision problem with particular importance for relative camera pose estimation and structure-from-motion. Existing local feature matching approaches, trained with correspondence supervision on large-scale datasets, obtain highly-accurate matches on the test sets. However, they do not generalise well to new datasets with different characteristics to those they were trained on, unlike classic feature extractors. Instead, they require finetuning, which assumes that ground-truth correspondences or ground-truth camera poses and 3D structure are available. We relax this assumption by removing the requirement of 3D structure, e.g., depth maps or point clouds, and only require camera pose information, which can be obtained from odometry. We do so by replacing correspondence losses with epipolar losses, which encourage putative matches to lie on the associated epipolar line. While weaker than correspondence supervision, we observe that this cue is sufficient for finetuning existing models on new data. We then further relax the assumption of known camera poses by using pose estimates in a novel bootstrapping approach. We evaluate on highly challenging datasets, including an indoor drone dataset and an outdoor smartphone camera dataset, and obtain state-of-the-art results without strong supervision.

Original languageEnglish
Title of host publicationProceedings - 2024 International Conference on 3D Vision, 3DV 2024
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages21-30
Number of pages10
ISBN (Electronic)9798350362459
ISBN (Print)979-8-3503-6246-6
DOIs
Publication statusPublished - 12 Jun 2024
Event11th International Conference on 3D Vision, 3DV 2024 - Davos, Switzerland
Duration: 18 Mar 202421 Mar 2024

Publication series

NameProceedings - 2024 International Conference on 3D Vision, 3DV 2024

Conference

Conference11th International Conference on 3D Vision, 3DV 2024
Country/TerritorySwitzerland
CityDavos
Period18/03/2421/03/24

Fingerprint

Dive into the research topics of 'SCENES: Subpixel Correspondence Estimation With Epipolar Supervision'. Together they form a unique fingerprint.

Cite this