Multidirectional Conjugate Gradients for Scalable Bundle Adjustment

Simon Weber, Nikolaus Demmel, Daniel Cremers

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

2 Scopus citations

Abstract

We revisit the problem of large-scale bundle adjustment and propose a technique called Multidirectional Conjugate Gradients that accelerates the solution of the normal equation by up to 61%. The key idea is that we enlarge the search space of classical preconditioned conjugate gradients to include multiple search directions. As a consequence, the resulting algorithm requires fewer iterations, leading to a significant speedup of large-scale reconstruction, in particular for denser problems where traditional approaches notoriously struggle. We provide a number of experimental ablation studies revealing the robustness to variations in the hyper-parameters and the speedup as a function of problem density.

Original languageEnglish
Title of host publicationPattern Recognition - 43rd DAGM German Conference, DAGM GCPR 2021, Proceedings
EditorsChristian Bauckhage, Juergen Gall, Alexander Schwing
PublisherSpringer Science and Business Media Deutschland GmbH
Pages712-724
Number of pages13
ISBN (Print)9783030926588
DOIs
StatePublished - 2021
Event43rd DAGM German Conference on Pattern Recognition, DAGM GCPR 2021 - Virtual, Online
Duration: 28 Sep 20211 Oct 2021

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume13024 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference43rd DAGM German Conference on Pattern Recognition, DAGM GCPR 2021
CityVirtual, Online
Period28/09/211/10/21

Keywords

  • Bundle adjustment
  • Large-scale reconstruction
  • Preconditioned conjugate gradients

Fingerprint

Dive into the research topics of 'Multidirectional Conjugate Gradients for Scalable Bundle Adjustment'. Together they form a unique fingerprint.

Cite this