Single image 3D human pose estimation from noisy observations

Edgar Simo Serra, A. Ramisa, G. Alenya, C. Torras, F. Moreno-Noguer

Research output: Chapter in Book/Report/Conference proceedingConference contribution

70 Citations (Scopus)

Abstract

Markerless 3D human pose detection from a single image is a severely underconstrained problem because different 3D poses can have similar image projections. In order to handle this ambiguity, current approaches rely on prior shape models that can only be correctly adjusted if 2D image features are accurately detected. Unfortunately, although current 2D part detector algorithms have shown promising results, they are not yet accurate enough to guarantee a complete disambiguation of the 3D inferred shape. In this paper, we introduce a novel approach for estimating 3D human pose even when observations are noisy. We propose a stochastic sampling strategy to propagate the noise from the image plane to the shape space. This provides a set of ambiguous 3D shapes, which are virtually undistinguishable from their image projections. Disambiguation is then achieved by imposing kinematic constraints that guarantee the resulting pose resembles a 3D human shape. We validate the method on a variety of situations in which state-of-the-art 2D detectors yield either inaccurate estimations or partly miss some of the body parts.

Original languageEnglish
Title of host publication2012 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2012
Pages2673-2680
Number of pages8
DOIs
Publication statusPublished - 2012
Externally publishedYes
Event2012 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2012 - Providence, RI, United States
Duration: 2012 Jun 162012 Jun 21

Other

Other2012 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2012
CountryUnited States
CityProvidence, RI
Period12/6/1612/6/21

Fingerprint

Detectors
Kinematics
Sampling

ASJC Scopus subject areas

  • Software
  • Computer Vision and Pattern Recognition

Cite this

Simo Serra, E., Ramisa, A., Alenya, G., Torras, C., & Moreno-Noguer, F. (2012). Single image 3D human pose estimation from noisy observations. In 2012 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2012 (pp. 2673-2680). [6247988] https://doi.org/10.1109/CVPR.2012.6247988

Single image 3D human pose estimation from noisy observations. / Simo Serra, Edgar; Ramisa, A.; Alenya, G.; Torras, C.; Moreno-Noguer, F.

2012 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2012. 2012. p. 2673-2680 6247988.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Simo Serra, E, Ramisa, A, Alenya, G, Torras, C & Moreno-Noguer, F 2012, Single image 3D human pose estimation from noisy observations. in 2012 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2012., 6247988, pp. 2673-2680, 2012 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2012, Providence, RI, United States, 12/6/16. https://doi.org/10.1109/CVPR.2012.6247988
Simo Serra E, Ramisa A, Alenya G, Torras C, Moreno-Noguer F. Single image 3D human pose estimation from noisy observations. In 2012 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2012. 2012. p. 2673-2680. 6247988 https://doi.org/10.1109/CVPR.2012.6247988
Simo Serra, Edgar ; Ramisa, A. ; Alenya, G. ; Torras, C. ; Moreno-Noguer, F. / Single image 3D human pose estimation from noisy observations. 2012 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2012. 2012. pp. 2673-2680
@inproceedings{6ae5121bcadc4cef82c090453bc1ef49,
title = "Single image 3D human pose estimation from noisy observations",
abstract = "Markerless 3D human pose detection from a single image is a severely underconstrained problem because different 3D poses can have similar image projections. In order to handle this ambiguity, current approaches rely on prior shape models that can only be correctly adjusted if 2D image features are accurately detected. Unfortunately, although current 2D part detector algorithms have shown promising results, they are not yet accurate enough to guarantee a complete disambiguation of the 3D inferred shape. In this paper, we introduce a novel approach for estimating 3D human pose even when observations are noisy. We propose a stochastic sampling strategy to propagate the noise from the image plane to the shape space. This provides a set of ambiguous 3D shapes, which are virtually undistinguishable from their image projections. Disambiguation is then achieved by imposing kinematic constraints that guarantee the resulting pose resembles a 3D human shape. We validate the method on a variety of situations in which state-of-the-art 2D detectors yield either inaccurate estimations or partly miss some of the body parts.",
author = "{Simo Serra}, Edgar and A. Ramisa and G. Alenya and C. Torras and F. Moreno-Noguer",
year = "2012",
doi = "10.1109/CVPR.2012.6247988",
language = "English",
isbn = "9781467312264",
pages = "2673--2680",
booktitle = "2012 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2012",

}

TY - GEN

T1 - Single image 3D human pose estimation from noisy observations

AU - Simo Serra, Edgar

AU - Ramisa, A.

AU - Alenya, G.

AU - Torras, C.

AU - Moreno-Noguer, F.

PY - 2012

Y1 - 2012

N2 - Markerless 3D human pose detection from a single image is a severely underconstrained problem because different 3D poses can have similar image projections. In order to handle this ambiguity, current approaches rely on prior shape models that can only be correctly adjusted if 2D image features are accurately detected. Unfortunately, although current 2D part detector algorithms have shown promising results, they are not yet accurate enough to guarantee a complete disambiguation of the 3D inferred shape. In this paper, we introduce a novel approach for estimating 3D human pose even when observations are noisy. We propose a stochastic sampling strategy to propagate the noise from the image plane to the shape space. This provides a set of ambiguous 3D shapes, which are virtually undistinguishable from their image projections. Disambiguation is then achieved by imposing kinematic constraints that guarantee the resulting pose resembles a 3D human shape. We validate the method on a variety of situations in which state-of-the-art 2D detectors yield either inaccurate estimations or partly miss some of the body parts.

AB - Markerless 3D human pose detection from a single image is a severely underconstrained problem because different 3D poses can have similar image projections. In order to handle this ambiguity, current approaches rely on prior shape models that can only be correctly adjusted if 2D image features are accurately detected. Unfortunately, although current 2D part detector algorithms have shown promising results, they are not yet accurate enough to guarantee a complete disambiguation of the 3D inferred shape. In this paper, we introduce a novel approach for estimating 3D human pose even when observations are noisy. We propose a stochastic sampling strategy to propagate the noise from the image plane to the shape space. This provides a set of ambiguous 3D shapes, which are virtually undistinguishable from their image projections. Disambiguation is then achieved by imposing kinematic constraints that guarantee the resulting pose resembles a 3D human shape. We validate the method on a variety of situations in which state-of-the-art 2D detectors yield either inaccurate estimations or partly miss some of the body parts.

UR - http://www.scopus.com/inward/record.url?scp=84866718027&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84866718027&partnerID=8YFLogxK

U2 - 10.1109/CVPR.2012.6247988

DO - 10.1109/CVPR.2012.6247988

M3 - Conference contribution

AN - SCOPUS:84866718027

SN - 9781467312264

SP - 2673

EP - 2680

BT - 2012 IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2012

ER -