Sequential Gaussian mixture models for two-level conditional random fields

Research output: Contribution to journal › Conference article › Research › peer review

Authors

  • Sergey Kosov
  • Franz Rottensteiner
  • Christian Heipke

Research Organisations

View graph of relations

Details

Original languageEnglish
Pages (from-to)153-163
Number of pages11
JournalLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume8142 LNCS
Publication statusPublished - 2013
Event35th German Conference on Pattern Recognition, GCPR 2013 - Saarbrücken, Germany
Duration: 3 Sept 2013 → 6 Sept 2013

Abstract

Conditional Random Fields are among the most popular techniques for image labelling because of their flexibility in modelling dependencies between the labels and the image features. This paper addresses the problem of efficient classification of partially occluded objects. For this purpose we propose a novel Gaussian Mixture Model based on a sequential training procedure, in combination with multi-level CRF-framework. Our approach is evaluated on urban aerial images. It is shown to increase the classification accuracy in occluded areas by up to 14,4%.

ASJC Scopus subject areas

Cite this

Sequential Gaussian mixture models for two-level conditional random fields. / Kosov, Sergey; Rottensteiner, Franz; Heipke, Christian.
In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Vol. 8142 LNCS, 2013, p. 153-163.

Research output: Contribution to journal › Conference article › Research › peer review

Kosov, S, Rottensteiner, F & Heipke, C 2013, 'Sequential Gaussian mixture models for two-level conditional random fields', Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 8142 LNCS, pp. 153-163. https://doi.org/10.1007/978-3-642-40602-7_16
Kosov, S., Rottensteiner, F., & Heipke, C. (2013). Sequential Gaussian mixture models for two-level conditional random fields. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 8142 LNCS, 153-163. https://doi.org/10.1007/978-3-642-40602-7_16
Kosov S, Rottensteiner F, Heipke C. Sequential Gaussian mixture models for two-level conditional random fields. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2013;8142 LNCS:153-163. doi: 10.1007/978-3-642-40602-7_16
Kosov, Sergey ; Rottensteiner, Franz ; Heipke, Christian. / Sequential Gaussian mixture models for two-level conditional random fields. In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2013 ; Vol. 8142 LNCS. pp. 153-163.
Download
@article{d04fbcc711604389b27c35a851c92d42,
title = "Sequential Gaussian mixture models for two-level conditional random fields",
abstract = "Conditional Random Fields are among the most popular techniques for image labelling because of their flexibility in modelling dependencies between the labels and the image features. This paper addresses the problem of efficient classification of partially occluded objects. For this purpose we propose a novel Gaussian Mixture Model based on a sequential training procedure, in combination with multi-level CRF-framework. Our approach is evaluated on urban aerial images. It is shown to increase the classification accuracy in occluded areas by up to 14,4%.",
author = "Sergey Kosov and Franz Rottensteiner and Christian Heipke",
year = "2013",
doi = "10.1007/978-3-642-40602-7_16",
language = "English",
volume = "8142 LNCS",
pages = "153--163",
note = "35th German Conference on Pattern Recognition, GCPR 2013 ; Conference date: 03-09-2013 Through 06-09-2013",

}

Download

TY - JOUR

T1 - Sequential Gaussian mixture models for two-level conditional random fields

AU - Kosov, Sergey

AU - Rottensteiner, Franz

AU - Heipke, Christian

PY - 2013

Y1 - 2013

N2 - Conditional Random Fields are among the most popular techniques for image labelling because of their flexibility in modelling dependencies between the labels and the image features. This paper addresses the problem of efficient classification of partially occluded objects. For this purpose we propose a novel Gaussian Mixture Model based on a sequential training procedure, in combination with multi-level CRF-framework. Our approach is evaluated on urban aerial images. It is shown to increase the classification accuracy in occluded areas by up to 14,4%.

AB - Conditional Random Fields are among the most popular techniques for image labelling because of their flexibility in modelling dependencies between the labels and the image features. This paper addresses the problem of efficient classification of partially occluded objects. For this purpose we propose a novel Gaussian Mixture Model based on a sequential training procedure, in combination with multi-level CRF-framework. Our approach is evaluated on urban aerial images. It is shown to increase the classification accuracy in occluded areas by up to 14,4%.

UR - http://www.scopus.com/inward/record.url?scp=84886427489&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-40602-7_16

DO - 10.1007/978-3-642-40602-7_16

M3 - Conference article

AN - SCOPUS:84886427489

VL - 8142 LNCS

SP - 153

EP - 163

JO - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

JF - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SN - 0302-9743

T2 - 35th German Conference on Pattern Recognition, GCPR 2013

Y2 - 3 September 2013 through 6 September 2013

ER -