TY - GEN
T1 - Parametric image segmentation of humans with structural shape priors
AU - Popa, Alin Ionut
AU - Sminchisescu, Cristian
PY - 2017
Y1 - 2017
N2 - 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.
AB - 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.
UR - http://www.scopus.com/inward/record.url?scp=85016190750&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-54184-6_5
DO - 10.1007/978-3-319-54184-6_5
M3 - Paper in conference proceeding
AN - SCOPUS:85016190750
SN - 9783319541839
VL - 10112 LNCS
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 68
EP - 83
BT - Computer Vision - ACCV 2016 - 13th Asian Conference on Computer Vision, Revised Selected Papers
PB - Springer
T2 - 13th Asian Conference on Computer Vision, ACCV 2016
Y2 - 20 November 2016 through 24 November 2016
ER -