Human parsing with a cascade of hierarchical poselet based pruners

Duan Tran, Yang Wang, David Alexander Forsyth

Research output: Contribution to journalConference article

Abstract

We address the problem of human parsing using part-based models. In particular, we consider part-based models that exploit rich pairwise relationship between parts, e.g. the color symmetry between left/right limbs. This poses a computational challenge since the state space of each part is very large, and algorithmic tricks (e.g. the distance transform) cannot be applied to handle these types of pairwise relationships. We propose to prune the state space of each part using a cascade of pruners. These pruners can filter out 99.6% of the states per part to about 500 states per part, while keeping the ground-truth states in the pruned state most of the time. In the pruned space, we can afford to apply human parsing models with more complex pairwise relationships between parts, such as the color symmetry. We demonstrate our method on a challenging human parsing dataset.

Original languageEnglish (US)
Article number6890316
JournalProceedings - IEEE International Conference on Multimedia and Expo
Volume2014-September
Issue numberSeptmber
DOIs
StatePublished - Sep 3 2014
Event2014 IEEE International Conference on Multimedia and Expo, ICME 2014 - Chengdu, China
Duration: Jul 14 2014Jul 18 2014

Fingerprint

Color
Ground state

Keywords

  • gesture analysis
  • human pose estimation

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Computer Science Applications

Cite this

Human parsing with a cascade of hierarchical poselet based pruners. / Tran, Duan; Wang, Yang; Forsyth, David Alexander.

In: Proceedings - IEEE International Conference on Multimedia and Expo, Vol. 2014-September, No. Septmber, 6890316, 03.09.2014.

Research output: Contribution to journalConference article

@article{0ea17c5cc39542f291fa87ad658c79b5,
title = "Human parsing with a cascade of hierarchical poselet based pruners",
abstract = "We address the problem of human parsing using part-based models. In particular, we consider part-based models that exploit rich pairwise relationship between parts, e.g. the color symmetry between left/right limbs. This poses a computational challenge since the state space of each part is very large, and algorithmic tricks (e.g. the distance transform) cannot be applied to handle these types of pairwise relationships. We propose to prune the state space of each part using a cascade of pruners. These pruners can filter out 99.6{\%} of the states per part to about 500 states per part, while keeping the ground-truth states in the pruned state most of the time. In the pruned space, we can afford to apply human parsing models with more complex pairwise relationships between parts, such as the color symmetry. We demonstrate our method on a challenging human parsing dataset.",
keywords = "gesture analysis, human pose estimation",
author = "Duan Tran and Yang Wang and Forsyth, {David Alexander}",
year = "2014",
month = "9",
day = "3",
doi = "10.1109/ICME.2014.6890316",
language = "English (US)",
volume = "2014-September",
journal = "Proceedings - IEEE International Conference on Multimedia and Expo",
issn = "1945-7871",
number = "Septmber",

}

TY - JOUR

T1 - Human parsing with a cascade of hierarchical poselet based pruners

AU - Tran, Duan

AU - Wang, Yang

AU - Forsyth, David Alexander

PY - 2014/9/3

Y1 - 2014/9/3

N2 - We address the problem of human parsing using part-based models. In particular, we consider part-based models that exploit rich pairwise relationship between parts, e.g. the color symmetry between left/right limbs. This poses a computational challenge since the state space of each part is very large, and algorithmic tricks (e.g. the distance transform) cannot be applied to handle these types of pairwise relationships. We propose to prune the state space of each part using a cascade of pruners. These pruners can filter out 99.6% of the states per part to about 500 states per part, while keeping the ground-truth states in the pruned state most of the time. In the pruned space, we can afford to apply human parsing models with more complex pairwise relationships between parts, such as the color symmetry. We demonstrate our method on a challenging human parsing dataset.

AB - We address the problem of human parsing using part-based models. In particular, we consider part-based models that exploit rich pairwise relationship between parts, e.g. the color symmetry between left/right limbs. This poses a computational challenge since the state space of each part is very large, and algorithmic tricks (e.g. the distance transform) cannot be applied to handle these types of pairwise relationships. We propose to prune the state space of each part using a cascade of pruners. These pruners can filter out 99.6% of the states per part to about 500 states per part, while keeping the ground-truth states in the pruned state most of the time. In the pruned space, we can afford to apply human parsing models with more complex pairwise relationships between parts, such as the color symmetry. We demonstrate our method on a challenging human parsing dataset.

KW - gesture analysis

KW - human pose estimation

UR - http://www.scopus.com/inward/record.url?scp=84937468150&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84937468150&partnerID=8YFLogxK

U2 - 10.1109/ICME.2014.6890316

DO - 10.1109/ICME.2014.6890316

M3 - Conference article

AN - SCOPUS:84937468150

VL - 2014-September

JO - Proceedings - IEEE International Conference on Multimedia and Expo

JF - Proceedings - IEEE International Conference on Multimedia and Expo

SN - 1945-7871

IS - Septmber

M1 - 6890316

ER -