Parametric image segmentation of humans with structural shape priors
(2017) 13th Asian Conference on Computer Vision, ACCV 2016 In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 10112 LNCS. p.68-83- Abstract
The figure-ground segmentation of humans in images captured in natural environments is an outstanding open problem due to the presence of complex backgrounds, articulation, varying body proportions, partial views and viewpoint changes. In this work we propose classspecific segmentation models that leverage parametric max-flow image segmentation and a large dataset of human shapes. Our contributions are as follows: (1) formulation of a submodular energy model that combines classspecific structural constraints and datadriven shape priors, within a parametric max-flow optimization methodology that systematically computes all breakpoints of the model in polynomial time; (2) design of a datadriven classspecific fusion methodology, based on... (More)
The figure-ground segmentation of humans in images captured in natural environments is an outstanding open problem due to the presence of complex backgrounds, articulation, varying body proportions, partial views and viewpoint changes. In this work we propose classspecific segmentation models that leverage parametric max-flow image segmentation and a large dataset of human shapes. Our contributions are as follows: (1) formulation of a submodular energy model that combines classspecific structural constraints and datadriven shape priors, within a parametric max-flow optimization methodology that systematically computes all breakpoints of the model in polynomial time; (2) design of a datadriven classspecific fusion methodology, based on matching against a large training set of exemplar human shapes (100,000 in our experiments), that allows the shape prior to be constructed on-the-fly, for arbitrary viewpoints and partial views.
(Less)
- author
- Popa, Alin Ionut and Sminchisescu, Cristian LU
- organization
- publishing date
- 2017
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- host publication
- Computer Vision - ACCV 2016 - 13th Asian Conference on Computer Vision, Revised Selected Papers
- series title
- Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
- volume
- 10112 LNCS
- pages
- 16 pages
- publisher
- Springer
- conference name
- 13th Asian Conference on Computer Vision, ACCV 2016
- conference location
- Taipei, Taiwan
- conference dates
- 2016-11-20 - 2016-11-24
- external identifiers
-
- scopus:85016190750
- ISSN
- 16113349
- 03029743
- ISBN
- 9783319541839
- DOI
- 10.1007/978-3-319-54184-6_5
- language
- English
- LU publication?
- yes
- id
- 391deb84-4f95-4ea1-b71e-a27ed4dcfbc8
- date added to LUP
- 2017-04-06 13:19:37
- date last changed
- 2025-01-07 10:55:37
@inproceedings{391deb84-4f95-4ea1-b71e-a27ed4dcfbc8, abstract = {{<p>The figure-ground segmentation of humans in images captured in natural environments is an outstanding open problem due to the presence of complex backgrounds, articulation, varying body proportions, partial views and viewpoint changes. In this work we propose classspecific segmentation models that leverage parametric max-flow image segmentation and a large dataset of human shapes. Our contributions are as follows: (1) formulation of a submodular energy model that combines classspecific structural constraints and datadriven shape priors, within a parametric max-flow optimization methodology that systematically computes all breakpoints of the model in polynomial time; (2) design of a datadriven classspecific fusion methodology, based on matching against a large training set of exemplar human shapes (100,000 in our experiments), that allows the shape prior to be constructed on-the-fly, for arbitrary viewpoints and partial views.</p>}}, author = {{Popa, Alin Ionut and Sminchisescu, Cristian}}, booktitle = {{Computer Vision - ACCV 2016 - 13th Asian Conference on Computer Vision, Revised Selected Papers}}, isbn = {{9783319541839}}, issn = {{16113349}}, language = {{eng}}, pages = {{68--83}}, publisher = {{Springer}}, series = {{Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)}}, title = {{Parametric image segmentation of humans with structural shape priors}}, url = {{http://dx.doi.org/10.1007/978-3-319-54184-6_5}}, doi = {{10.1007/978-3-319-54184-6_5}}, volume = {{10112 LNCS}}, year = {{2017}}, }