Deep fundamental matrix estimation without correspondences

Publikation: Bidrag til tidsskrift › Konferenceartikel › Forskning › fagfællebedømt

Standard

Deep fundamental matrix estimation without correspondences. / Poursaeed, Omid; Yang, Guandao; Prakash, Aditya; Fang, Qiuren; Jiang, Hanqing; Hariharan, Bharath; Belongie, Serge.

I: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2019, s. 485-497.

Publikation: Bidrag til tidsskrift › Konferenceartikel › Forskning › fagfællebedømt

Harvard

Poursaeed, O, Yang, G, Prakash, A, Fang, Q, Jiang, H, Hariharan, B & Belongie, S 2019, 'Deep fundamental matrix estimation without correspondences', Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), s. 485-497. https://doi.org/10.1007/978-3-030-11015-4_35

APA

Poursaeed, O., Yang, G., Prakash, A., Fang, Q., Jiang, H., Hariharan, B., & Belongie, S. (2019). Deep fundamental matrix estimation without correspondences. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 485-497. https://doi.org/10.1007/978-3-030-11015-4_35

Vancouver

Poursaeed O, Yang G, Prakash A, Fang Q, Jiang H, Hariharan B o.a. Deep fundamental matrix estimation without correspondences. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2019;485-497. https://doi.org/10.1007/978-3-030-11015-4_35

Author

Poursaeed, Omid ; Yang, Guandao ; Prakash, Aditya ; Fang, Qiuren ; Jiang, Hanqing ; Hariharan, Bharath ; Belongie, Serge. / Deep fundamental matrix estimation without correspondences. I: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2019 ; s. 485-497.

Bibtex

@inproceedings{59fb73d6b6cb4ffb9da492178a11d82c,
title = "Deep fundamental matrix estimation without correspondences",
abstract = "Estimating fundamental matrices is a classic problem in computer vision. Traditional methods rely heavily on the correctness of estimated key-point correspondences, which can be noisy and unreliable. As a result, it is difficult for these methods to handle image pairs with large occlusion or significantly different camera poses. In this paper, we propose novel neural network architectures to estimate fundamental matrices in an end-to-end manner without relying on point correspondences. New modules and layers are introduced in order to preserve mathematical properties of the fundamental matrix as a homogeneous rank-2 matrix with seven degrees of freedom. We analyze performance of the proposed models using various metrics on the KITTI dataset, and show that they achieve competitive performance with traditional methods without the need for extracting correspondences.",
keywords = "Deep learning, Epipolar geometry, Fundamental matrix, Stereo",
author = "Omid Poursaeed and Guandao Yang and Aditya Prakash and Qiuren Fang and Hanqing Jiang and Bharath Hariharan and Serge Belongie",
note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2019.; 15th European Conference on Computer Vision, ECCV 2018 ; Conference date: 08-09-2018 Through 14-09-2018",
year = "2019",
doi = "10.1007/978-3-030-11015-4_35",
language = "English",
pages = "485--497",
journal = "Lecture Notes in Computer Science",
issn = "0302-9743",
publisher = "Springer Verlag",

}

RIS

TY - GEN

T1 - Deep fundamental matrix estimation without correspondences

AU - Poursaeed, Omid

AU - Yang, Guandao

AU - Prakash, Aditya

AU - Fang, Qiuren

AU - Jiang, Hanqing

AU - Hariharan, Bharath

AU - Belongie, Serge

N1 - Publisher Copyright: © Springer Nature Switzerland AG 2019.

PY - 2019

Y1 - 2019

N2 - Estimating fundamental matrices is a classic problem in computer vision. Traditional methods rely heavily on the correctness of estimated key-point correspondences, which can be noisy and unreliable. As a result, it is difficult for these methods to handle image pairs with large occlusion or significantly different camera poses. In this paper, we propose novel neural network architectures to estimate fundamental matrices in an end-to-end manner without relying on point correspondences. New modules and layers are introduced in order to preserve mathematical properties of the fundamental matrix as a homogeneous rank-2 matrix with seven degrees of freedom. We analyze performance of the proposed models using various metrics on the KITTI dataset, and show that they achieve competitive performance with traditional methods without the need for extracting correspondences.

AB - Estimating fundamental matrices is a classic problem in computer vision. Traditional methods rely heavily on the correctness of estimated key-point correspondences, which can be noisy and unreliable. As a result, it is difficult for these methods to handle image pairs with large occlusion or significantly different camera poses. In this paper, we propose novel neural network architectures to estimate fundamental matrices in an end-to-end manner without relying on point correspondences. New modules and layers are introduced in order to preserve mathematical properties of the fundamental matrix as a homogeneous rank-2 matrix with seven degrees of freedom. We analyze performance of the proposed models using various metrics on the KITTI dataset, and show that they achieve competitive performance with traditional methods without the need for extracting correspondences.

KW - Deep learning

KW - Epipolar geometry

KW - Fundamental matrix

KW - Stereo

UR - http://www.scopus.com/inward/record.url?scp=85061693720&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-11015-4_35

DO - 10.1007/978-3-030-11015-4_35

M3 - Conference article

AN - SCOPUS:85061693720

SP - 485

EP - 497

JO - Lecture Notes in Computer Science

JF - Lecture Notes in Computer Science

SN - 0302-9743

T2 - 15th European Conference on Computer Vision, ECCV 2018

Y2 - 8 September 2018 through 14 September 2018

ER -

ID: 301824797