---
_id: '2171'
abstract:
- lang: eng
  text: We present LS-CRF, a new method for training cyclic Conditional Random Fields
    (CRFs) from large datasets that is inspired by classical closed-form expressions
    for the maximum likelihood parameters of a generative graphical model with tree
    topology. Training a CRF with LS-CRF requires only solving a set of independent
    regression problems, each of which can be solved efficiently in closed form or
    by an iterative solver. This makes LS-CRF orders of magnitude faster than classical
    CRF training based on probabilistic inference, and at the same time more flexible
    and easier to implement than other approximate techniques, such as pseudolikelihood
    or piecewise training. We apply LS-CRF to the task of semantic image segmentation,
    showing that it achieves on par accuracy to other training techniques at higher
    speed, thereby allowing efficient CRF training from very large training sets.
    For example, training a linearly parameterized pairwise CRF on 150,000 images
    requires less than one hour on a modern workstation.
alternative_title:
- LNCS
article_processing_charge: No
arxiv: 1
author:
- first_name: Alexander
  full_name: Kolesnikov, Alexander
  id: 2D157DB6-F248-11E8-B48F-1D18A9856A87
  last_name: Kolesnikov
- first_name: Matthieu
  full_name: Guillaumin, Matthieu
  last_name: Guillaumin
- first_name: Vittorio
  full_name: Ferrari, Vittorio
  last_name: Ferrari
- first_name: Christoph
  full_name: Lampert, Christoph
  id: 40C20FD2-F248-11E8-B48F-1D18A9856A87
  last_name: Lampert
  orcid: 0000-0001-8622-7887
citation:
  ama: 'Kolesnikov A, Guillaumin M, Ferrari V, Lampert C. Closed-form approximate
    CRF training for scalable image segmentation. In: Fleet D, Pajdla T, Schiele B,
    Tuytelaars T, eds. <i>Lecture Notes in Computer Science (Including Subseries Lecture
    Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i>. Vol
    8691. Springer; 2014:550-565. doi:<a href="https://doi.org/10.1007/978-3-319-10578-9_36">10.1007/978-3-319-10578-9_36</a>'
  apa: 'Kolesnikov, A., Guillaumin, M., Ferrari, V., &#38; Lampert, C. (2014). Closed-form
    approximate CRF training for scalable image segmentation. In D. Fleet, T. Pajdla,
    B. Schiele, &#38; T. Tuytelaars (Eds.), <i>Lecture Notes in Computer Science (including
    subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i>
    (Vol. 8691, pp. 550–565). Zurich, Switzerland: Springer. <a href="https://doi.org/10.1007/978-3-319-10578-9_36">https://doi.org/10.1007/978-3-319-10578-9_36</a>'
  chicago: Kolesnikov, Alexander, Matthieu Guillaumin, Vittorio Ferrari, and Christoph
    Lampert. “Closed-Form Approximate CRF Training for Scalable Image Segmentation.”
    In <i>Lecture Notes in Computer Science (Including Subseries Lecture Notes in
    Artificial Intelligence and Lecture Notes in Bioinformatics)</i>, edited by David
    Fleet, Tomas Pajdla, Bernt Schiele, and Tinne Tuytelaars, 8691:550–65. Springer,
    2014. <a href="https://doi.org/10.1007/978-3-319-10578-9_36">https://doi.org/10.1007/978-3-319-10578-9_36</a>.
  ieee: A. Kolesnikov, M. Guillaumin, V. Ferrari, and C. Lampert, “Closed-form approximate
    CRF training for scalable image segmentation,” in <i>Lecture Notes in Computer
    Science (including subseries Lecture Notes in Artificial Intelligence and Lecture
    Notes in Bioinformatics)</i>, Zurich, Switzerland, 2014, vol. 8691, no. PART 3,
    pp. 550–565.
  ista: 'Kolesnikov A, Guillaumin M, Ferrari V, Lampert C. 2014. Closed-form approximate
    CRF training for scalable image segmentation. Lecture Notes in Computer Science
    (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes
    in Bioinformatics). ECCV: European Conference on Computer Vision, LNCS, vol. 8691,
    550–565.'
  mla: Kolesnikov, Alexander, et al. “Closed-Form Approximate CRF Training for Scalable
    Image Segmentation.” <i>Lecture Notes in Computer Science (Including Subseries
    Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)</i>,
    edited by David Fleet et al., vol. 8691, no. PART 3, Springer, 2014, pp. 550–65,
    doi:<a href="https://doi.org/10.1007/978-3-319-10578-9_36">10.1007/978-3-319-10578-9_36</a>.
  short: A. Kolesnikov, M. Guillaumin, V. Ferrari, C. Lampert, in:, D. Fleet, T. Pajdla,
    B. Schiele, T. Tuytelaars (Eds.), Lecture Notes in Computer Science (Including
    Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics),
    Springer, 2014, pp. 550–565.
conference:
  end_date: 2014-09-12
  location: Zurich, Switzerland
  name: 'ECCV: European Conference on Computer Vision'
  start_date: 2014-09-06
date_created: 2018-12-11T11:56:07Z
date_published: 2014-09-01T00:00:00Z
date_updated: 2025-06-11T07:59:20Z
day: '01'
department:
- _id: ChLa
doi: 10.1007/978-3-319-10578-9_36
ec_funded: 1
editor:
- first_name: David
  full_name: Fleet, David
  last_name: Fleet
- first_name: Tomas
  full_name: Pajdla, Tomas
  last_name: Pajdla
- first_name: Bernt
  full_name: Schiele, Bernt
  last_name: Schiele
- first_name: Tinne
  full_name: Tuytelaars, Tinne
  last_name: Tuytelaars
external_id:
  arxiv:
  - '1403.7057'
intvolume: '      8691'
issue: PART 3
language:
- iso: eng
main_file_link:
- open_access: '1'
  url: http://arxiv.org/abs/1403.7057
month: '09'
oa: 1
oa_version: Submitted Version
page: 550 - 565
project:
- _id: 2532554C-B435-11E9-9278-68D0E5697425
  call_identifier: FP7
  grant_number: '308036'
  name: Lifelong Learning of Visual Scene Understanding
publication: Lecture Notes in Computer Science (including subseries Lecture Notes
  in Artificial Intelligence and Lecture Notes in Bioinformatics)
publication_status: published
publisher: Springer
publist_id: '4813'
quality_controlled: '1'
scopus_import: '1'
status: public
title: Closed-form approximate CRF training for scalable image segmentation
type: conference
user_id: 2DF688A6-F248-11E8-B48F-1D18A9856A87
volume: 8691
year: '2014'
...