Tracking people on a Torus

Ahmed Elgammal, Chan Su Lee

Research output: Contribution to journalArticle

65 Citations (Scopus)

Abstract

We present a framework for monocular 3D kinematic pose tracking and viewpoint estimation of periodic and quasi-periodic human motions from an uncalibrated camera. The approach we introduce here is based on learning both the visual observation manifold and the kinematic manifold of the motion using a joint representation. We show that the visual manifold of the observed shape of a human performing a periodic motion, observed from different viewpoints, is topologically equivalent to {\em a torus manifold}. The approach we introduce here is based on {\em supervised} learning of both the visual and kinematic manifolds. Instead of learning an embedding of the manifold, we learn the geometric deformation between an ideal manifold (conceptual equivalent topological structure) and a twisted version of the manifold (the data). Experimental results show accurate estimation of the 3D body posture and the viewpoint from a single uncalibrated camera.

Original languageEnglish (US)
Pages (from-to)520-538
Number of pages19
JournalIEEE Transactions on Pattern Analysis and Machine Intelligence
Volume31
Issue number3
DOIs
StatePublished - Feb 11 2009

Fingerprint

Torus
Kinematics
Cameras
Supervised learning
Camera
Motion
Periodic Motion
Topological Structure
Supervised Learning
Experimental Results
Vision

All Science Journal Classification (ASJC) codes

  • Software
  • Artificial Intelligence
  • Applied Mathematics
  • Computer Vision and Pattern Recognition
  • Computational Theory and Mathematics

Cite this

@article{d9304fb4749f487d855005b65d3db597,
title = "Tracking people on a Torus",
abstract = "We present a framework for monocular 3D kinematic pose tracking and viewpoint estimation of periodic and quasi-periodic human motions from an uncalibrated camera. The approach we introduce here is based on learning both the visual observation manifold and the kinematic manifold of the motion using a joint representation. We show that the visual manifold of the observed shape of a human performing a periodic motion, observed from different viewpoints, is topologically equivalent to {\em a torus manifold}. The approach we introduce here is based on {\em supervised} learning of both the visual and kinematic manifolds. Instead of learning an embedding of the manifold, we learn the geometric deformation between an ideal manifold (conceptual equivalent topological structure) and a twisted version of the manifold (the data). Experimental results show accurate estimation of the 3D body posture and the viewpoint from a single uncalibrated camera.",
author = "Ahmed Elgammal and Lee, {Chan Su}",
year = "2009",
month = "2",
day = "11",
doi = "https://doi.org/10.1109/TPAMI.2008.101",
language = "English (US)",
volume = "31",
pages = "520--538",
journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",
issn = "0162-8828",
publisher = "IEEE Computer Society",
number = "3",

}

Tracking people on a Torus. / Elgammal, Ahmed; Lee, Chan Su.

In: IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 31, No. 3, 11.02.2009, p. 520-538.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Tracking people on a Torus

AU - Elgammal, Ahmed

AU - Lee, Chan Su

PY - 2009/2/11

Y1 - 2009/2/11

N2 - We present a framework for monocular 3D kinematic pose tracking and viewpoint estimation of periodic and quasi-periodic human motions from an uncalibrated camera. The approach we introduce here is based on learning both the visual observation manifold and the kinematic manifold of the motion using a joint representation. We show that the visual manifold of the observed shape of a human performing a periodic motion, observed from different viewpoints, is topologically equivalent to {\em a torus manifold}. The approach we introduce here is based on {\em supervised} learning of both the visual and kinematic manifolds. Instead of learning an embedding of the manifold, we learn the geometric deformation between an ideal manifold (conceptual equivalent topological structure) and a twisted version of the manifold (the data). Experimental results show accurate estimation of the 3D body posture and the viewpoint from a single uncalibrated camera.

AB - We present a framework for monocular 3D kinematic pose tracking and viewpoint estimation of periodic and quasi-periodic human motions from an uncalibrated camera. The approach we introduce here is based on learning both the visual observation manifold and the kinematic manifold of the motion using a joint representation. We show that the visual manifold of the observed shape of a human performing a periodic motion, observed from different viewpoints, is topologically equivalent to {\em a torus manifold}. The approach we introduce here is based on {\em supervised} learning of both the visual and kinematic manifolds. Instead of learning an embedding of the manifold, we learn the geometric deformation between an ideal manifold (conceptual equivalent topological structure) and a twisted version of the manifold (the data). Experimental results show accurate estimation of the 3D body posture and the viewpoint from a single uncalibrated camera.

UR - http://www.scopus.com/inward/record.url?scp=59549087186&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=59549087186&partnerID=8YFLogxK

U2 - https://doi.org/10.1109/TPAMI.2008.101

DO - https://doi.org/10.1109/TPAMI.2008.101

M3 - Article

VL - 31

SP - 520

EP - 538

JO - IEEE Transactions on Pattern Analysis and Machine Intelligence

JF - IEEE Transactions on Pattern Analysis and Machine Intelligence

SN - 0162-8828

IS - 3

ER -