Joint optimisation for object class segmentation and dense stereo reconstruction

The problems of dense stereo reconstruction and object class segmentation can both be formulated as Conditional Random Field based labelling problems, in which every pixel in the image is assigned a label corresponding to either its disparity, or an object class such as road or building. While these...

Full description

Bibliographic Details
Main Authors: Ladický, L, Sturgess, P, Russell, C, Sengupta, S, Bastanlar, Y, Clocksin, W, Torr, PHS
Format: Conference item
Language:English
Published: British Machine Vision Association 2010
_version_ 1817931133293166592
author Ladický, L
Sturgess, P
Russell, C
Sengupta, S
Bastanlar, Y
Clocksin, W
Torr, PHS
author_facet Ladický, L
Sturgess, P
Russell, C
Sengupta, S
Bastanlar, Y
Clocksin, W
Torr, PHS
author_sort Ladický, L
collection OXFORD
description The problems of dense stereo reconstruction and object class segmentation can both be formulated as Conditional Random Field based labelling problems, in which every pixel in the image is assigned a label corresponding to either its disparity, or an object class such as road or building. While these two problems are mutually informative, no attempt has been made to jointly optimise their labellings. In this work we provide a principled energy minimisation framework that unifies the two problems and demonstrate that, by resolving ambiguities in real world data, joint optimisation of the two problems substantially improves performance. To evaluate our method, we augment the street view Leuven data set, producing 70 hand labelled object class and disparity maps. We hope that the release of these annotations will stimulate further work in the challenging domain of street-view analysis.
first_indexed 2024-12-09T03:17:10Z
format Conference item
id oxford-uuid:a75c41c6-cd2b-4aca-8272-57d36f7a8a8e
institution University of Oxford
language English
last_indexed 2024-12-09T03:17:10Z
publishDate 2010
publisher British Machine Vision Association
record_format dspace
spelling oxford-uuid:a75c41c6-cd2b-4aca-8272-57d36f7a8a8e2024-10-24T14:59:09ZJoint optimisation for object class segmentation and dense stereo reconstructionConference itemhttp://purl.org/coar/resource_type/c_5794uuid:a75c41c6-cd2b-4aca-8272-57d36f7a8a8eEnglishSymplectic ElementsBritish Machine Vision Association2010Ladický, LSturgess, PRussell, CSengupta, SBastanlar, YClocksin, WTorr, PHSThe problems of dense stereo reconstruction and object class segmentation can both be formulated as Conditional Random Field based labelling problems, in which every pixel in the image is assigned a label corresponding to either its disparity, or an object class such as road or building. While these two problems are mutually informative, no attempt has been made to jointly optimise their labellings. In this work we provide a principled energy minimisation framework that unifies the two problems and demonstrate that, by resolving ambiguities in real world data, joint optimisation of the two problems substantially improves performance. To evaluate our method, we augment the street view Leuven data set, producing 70 hand labelled object class and disparity maps. We hope that the release of these annotations will stimulate further work in the challenging domain of street-view analysis.
spellingShingle Ladický, L
Sturgess, P
Russell, C
Sengupta, S
Bastanlar, Y
Clocksin, W
Torr, PHS
Joint optimisation for object class segmentation and dense stereo reconstruction
title Joint optimisation for object class segmentation and dense stereo reconstruction
title_full Joint optimisation for object class segmentation and dense stereo reconstruction
title_fullStr Joint optimisation for object class segmentation and dense stereo reconstruction
title_full_unstemmed Joint optimisation for object class segmentation and dense stereo reconstruction
title_short Joint optimisation for object class segmentation and dense stereo reconstruction
title_sort joint optimisation for object class segmentation and dense stereo reconstruction
work_keys_str_mv AT ladickyl jointoptimisationforobjectclasssegmentationanddensestereoreconstruction
AT sturgessp jointoptimisationforobjectclasssegmentationanddensestereoreconstruction
AT russellc jointoptimisationforobjectclasssegmentationanddensestereoreconstruction
AT senguptas jointoptimisationforobjectclasssegmentationanddensestereoreconstruction
AT bastanlary jointoptimisationforobjectclasssegmentationanddensestereoreconstruction
AT clocksinw jointoptimisationforobjectclasssegmentationanddensestereoreconstruction
AT torrphs jointoptimisationforobjectclasssegmentationanddensestereoreconstruction