Learning to solve nonlinear least squares for monocular stereo

File Description SizeFormat 
Learning_to_Solve__ECCV_camera_ready.pdfFile embargoed until 07 October 20192.11 MBAdobe PDF    Request a copy
Title: Learning to solve nonlinear least squares for monocular stereo
Authors: Clark, R
Bloesch, M
Czarnowski, J
Leutenegger, S
Davison, AJ
Item Type: Conference Paper
Abstract: Sum-of-squares objective functions are very popular in computer vision algorithms. However, these objective functions are not always easy to optimize. The underlying assumptions made by solvers are often not satisfied and many problems are inherently ill-posed. In this paper, we propose a neural nonlinear least squares optimization algorithm which learns to effectively optimize these cost functions even in the presence of adversities. Unlike traditional approaches, the proposed solver requires no hand-crafted regularizers or priors as these are implicitly learned from the data. We apply our method to the problem of motion stereo ie. jointly estimating the motion and scene geometry from pairs of images of a monocular sequence. We show that our learned optimizer is able to efficiently and effectively solve this challenging optimization problem.
Issue Date: 7-Oct-2018
Date of Acceptance: 8-Sep-2018
URI: http://hdl.handle.net/10044/1/64919
DOI: https://dx.doi.org/10.1007/978-3-030-01237-3_18
ISBN: 9783030012366
ISSN: 0302-9743
Publisher: Springer Nature Switzerland AG 2018
Start Page: 291
End Page: 306
Journal / Book Title: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume: 11212 LNCS
Copyright Statement: © 2018 Springer-Verlag. The final publication is available at Springer via https://dx.doi.org/10.1007/978-3-030-01237-3_18.
Sponsor/Funder: Dyson Technology Limited
Funder's Grant Number: PO 4500421830
Conference Name: 15th European Conference on Computer Vision
Keywords: 08 Information And Computing Sciences
Artificial Intelligence & Image Processing
Publication Status: Published
Start Date: 2018-09-08
Finish Date: 2018-09-14
Conference Place: Munich, Germany
Embargo Date: 2019-10-07
Online Publication Date: 2018-10-07
Appears in Collections:Faculty of Engineering
Computing



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Creative Commonsx