Estimating 2D Upper Body Poses from Monocular Images

Jeroen Broekhuijsen, Ronald Walter Poppe, Mannes Poel

Abstract

Automatic estimation and recognition of poses from video allows for a whole range of applications. The research described here is an important step towards automatic extraction of 3D poses. We describe our research to extract the 2D joint locations of the people in meeting videos. The key point of the research described here is that we generalize over variations in appearance of both people and scene. This results in a robust detection of 2D joint locations. For the detection of different limbs, we employ a number of limb locators. Each of these uses a different set of image features. We evaluate our work on two videos that have been recorded in the meeting context. Our results are promising, yielding an average error of approximately 3-5 cm per joint.
Original languageUndefined
Place of PublicationEnschede
PublisherCentrum voor Telematica en Informatie Technologie
Number of pages12
StatePublished - 6 Sep 2006

Publication series

NameCTIT Technical Report Series
PublisherUniversity of Twente, Centre for Telematics and Information Technology
No.06-55
ISSN (Print)1381-3625

Keywords

  • EWI-6903
  • EC Grant Agreement nr.: FP6/506811
  • IR-66349
  • METIS-238676

Cite this

Broekhuijsen, J., Poppe, R. W., & Poel, M. (2006). Estimating 2D Upper Body Poses from Monocular Images. (CTIT Technical Report Series; No. 06-55). Enschede: Centrum voor Telematica en Informatie Technologie.

Broekhuijsen, Jeroen; Poppe, Ronald Walter; Poel, Mannes / Estimating 2D Upper Body Poses from Monocular Images.

Enschede : Centrum voor Telematica en Informatie Technologie, 2006. 12 p. (CTIT Technical Report Series; No. 06-55).

Research output: ProfessionalReport

@book{af9bf7afbf8643cf9a6db4baf6c3ff99,
title = "Estimating 2D Upper Body Poses from Monocular Images",
abstract = "Automatic estimation and recognition of poses from video allows for a whole range of applications. The research described here is an important step towards automatic extraction of 3D poses. We describe our research to extract the 2D joint locations of the people in meeting videos. The key point of the research described here is that we generalize over variations in appearance of both people and scene. This results in a robust detection of 2D joint locations. For the detection of different limbs, we employ a number of limb locators. Each of these uses a different set of image features. We evaluate our work on two videos that have been recorded in the meeting context. Our results are promising, yielding an average error of approximately 3-5 cm per joint.",
keywords = "EWI-6903, EC Grant Agreement nr.: FP6/506811, IR-66349, METIS-238676",
author = "Jeroen Broekhuijsen and Poppe, {Ronald Walter} and Mannes Poel",
year = "2006",
month = "9",
series = "CTIT Technical Report Series",
publisher = "Centrum voor Telematica en Informatie Technologie",
number = "06-55",

}

Broekhuijsen, J, Poppe, RW & Poel, M 2006, Estimating 2D Upper Body Poses from Monocular Images. CTIT Technical Report Series, no. 06-55, Centrum voor Telematica en Informatie Technologie, Enschede.

Estimating 2D Upper Body Poses from Monocular Images. / Broekhuijsen, Jeroen; Poppe, Ronald Walter; Poel, Mannes.

Enschede : Centrum voor Telematica en Informatie Technologie, 2006. 12 p. (CTIT Technical Report Series; No. 06-55).

Research output: ProfessionalReport

TY - BOOK

T1 - Estimating 2D Upper Body Poses from Monocular Images

AU - Broekhuijsen,Jeroen

AU - Poppe,Ronald Walter

AU - Poel,Mannes

PY - 2006/9/6

Y1 - 2006/9/6

N2 - Automatic estimation and recognition of poses from video allows for a whole range of applications. The research described here is an important step towards automatic extraction of 3D poses. We describe our research to extract the 2D joint locations of the people in meeting videos. The key point of the research described here is that we generalize over variations in appearance of both people and scene. This results in a robust detection of 2D joint locations. For the detection of different limbs, we employ a number of limb locators. Each of these uses a different set of image features. We evaluate our work on two videos that have been recorded in the meeting context. Our results are promising, yielding an average error of approximately 3-5 cm per joint.

AB - Automatic estimation and recognition of poses from video allows for a whole range of applications. The research described here is an important step towards automatic extraction of 3D poses. We describe our research to extract the 2D joint locations of the people in meeting videos. The key point of the research described here is that we generalize over variations in appearance of both people and scene. This results in a robust detection of 2D joint locations. For the detection of different limbs, we employ a number of limb locators. Each of these uses a different set of image features. We evaluate our work on two videos that have been recorded in the meeting context. Our results are promising, yielding an average error of approximately 3-5 cm per joint.

KW - EWI-6903

KW - EC Grant Agreement nr.: FP6/506811

KW - IR-66349

KW - METIS-238676

M3 - Report

T3 - CTIT Technical Report Series

BT - Estimating 2D Upper Body Poses from Monocular Images

PB - Centrum voor Telematica en Informatie Technologie

ER -

Broekhuijsen J, Poppe RW, Poel M. Estimating 2D Upper Body Poses from Monocular Images. Enschede: Centrum voor Telematica en Informatie Technologie, 2006. 12 p. (CTIT Technical Report Series; 06-55).