Shape-constrained Gaussian Process Regression for Facial-point-based Head-pose Normalization

Ognjen Rudovic, Maja Pantic

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

11 Citations (Scopus)

Abstract

Given the facial points extracted from an image of a face in an arbitrary pose, the goal of facial-point-based headpose normalization is to obtain the corresponding facial points in a predefined pose (e.g., frontal). This involves inference of complex and high-dimensional mappings due to the large number of the facial points employed, and due to differences in head-pose and facial expression. Most regression-based approaches for learning such mappings focus on modeling correlations only between the inputs (i.e., the facial points in a non-frontal pose) and the outputs (i.e., the facial points in the frontal pose), but not within the inputs and the outputs of the model. This makes these models prone to errors due to noise and outliers in test data, often resulting in anatomically impossible facial configurations formed by their predictions. To address this, we propose Shape-constrained Gaussian Process (SC-GP) regression for facial-point-based head-pose normalization. Specifically, a deformable face-shape model is used to learn a face-shape prior, which is placed on both the input and the output of GP regression in order to constrain the model predictions to anatomically feasible facial configurations. Our extensive experiments on both synthetic and real image data show that the proposed approach generalizes well across poses and handles successfully noise and outliers in test data. In addition, the proposed model outperforms previously proposed approaches to facial-point-based head-pose normalization.
Original languageUndefined
Title of host publicationIEEE International Conference on Computer Vision, ICCV 2011
Place of PublicationUSA
PublisherIEEE Computer Society
Pages1495-1502
Number of pages8
ISBN (Print)978-1-4577-1101-5
DOIs
Publication statusPublished - Nov 2011
EventIEEE International Conference on Computer Vision 2011 - Fira de Barcelona, Barcelona, Spain
Duration: 6 Nov 201113 Nov 2011

Publication series

Name
PublisherIEEE Computer Society
ISSN (Print)1550-5499

Conference

ConferenceIEEE International Conference on Computer Vision 2011
Abbreviated titleICCV 2011
CountrySpain
CityBarcelona
Period6/11/1113/11/11

Keywords

  • METIS-285022
  • IR-79458
  • EWI-21316
  • HMI-MI: MULTIMODAL INTERACTIONS
  • EC Grant Agreement nr.: ERC/203143

Cite this

Rudovic, O., & Pantic, M. (2011). Shape-constrained Gaussian Process Regression for Facial-point-based Head-pose Normalization. In IEEE International Conference on Computer Vision, ICCV 2011 (pp. 1495-1502). USA: IEEE Computer Society. https://doi.org/10.1109/ICCV.2011.6126407
Rudovic, Ognjen ; Pantic, Maja. / Shape-constrained Gaussian Process Regression for Facial-point-based Head-pose Normalization. IEEE International Conference on Computer Vision, ICCV 2011. USA : IEEE Computer Society, 2011. pp. 1495-1502
@inproceedings{2c7e427c02fa4b2bbcabbf3cb3fe37d2,
title = "Shape-constrained Gaussian Process Regression for Facial-point-based Head-pose Normalization",
abstract = "Given the facial points extracted from an image of a face in an arbitrary pose, the goal of facial-point-based headpose normalization is to obtain the corresponding facial points in a predefined pose (e.g., frontal). This involves inference of complex and high-dimensional mappings due to the large number of the facial points employed, and due to differences in head-pose and facial expression. Most regression-based approaches for learning such mappings focus on modeling correlations only between the inputs (i.e., the facial points in a non-frontal pose) and the outputs (i.e., the facial points in the frontal pose), but not within the inputs and the outputs of the model. This makes these models prone to errors due to noise and outliers in test data, often resulting in anatomically impossible facial configurations formed by their predictions. To address this, we propose Shape-constrained Gaussian Process (SC-GP) regression for facial-point-based head-pose normalization. Specifically, a deformable face-shape model is used to learn a face-shape prior, which is placed on both the input and the output of GP regression in order to constrain the model predictions to anatomically feasible facial configurations. Our extensive experiments on both synthetic and real image data show that the proposed approach generalizes well across poses and handles successfully noise and outliers in test data. In addition, the proposed model outperforms previously proposed approaches to facial-point-based head-pose normalization.",
keywords = "METIS-285022, IR-79458, EWI-21316, HMI-MI: MULTIMODAL INTERACTIONS, EC Grant Agreement nr.: ERC/203143",
author = "Ognjen Rudovic and Maja Pantic",
note = "eemcs-eprint-21316",
year = "2011",
month = "11",
doi = "10.1109/ICCV.2011.6126407",
language = "Undefined",
isbn = "978-1-4577-1101-5",
publisher = "IEEE Computer Society",
pages = "1495--1502",
booktitle = "IEEE International Conference on Computer Vision, ICCV 2011",
address = "United States",

}

Rudovic, O & Pantic, M 2011, Shape-constrained Gaussian Process Regression for Facial-point-based Head-pose Normalization. in IEEE International Conference on Computer Vision, ICCV 2011. IEEE Computer Society, USA, pp. 1495-1502, IEEE International Conference on Computer Vision 2011, Barcelona, Spain, 6/11/11. https://doi.org/10.1109/ICCV.2011.6126407

Shape-constrained Gaussian Process Regression for Facial-point-based Head-pose Normalization. / Rudovic, Ognjen; Pantic, Maja.

IEEE International Conference on Computer Vision, ICCV 2011. USA : IEEE Computer Society, 2011. p. 1495-1502.

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - Shape-constrained Gaussian Process Regression for Facial-point-based Head-pose Normalization

AU - Rudovic, Ognjen

AU - Pantic, Maja

N1 - eemcs-eprint-21316

PY - 2011/11

Y1 - 2011/11

N2 - Given the facial points extracted from an image of a face in an arbitrary pose, the goal of facial-point-based headpose normalization is to obtain the corresponding facial points in a predefined pose (e.g., frontal). This involves inference of complex and high-dimensional mappings due to the large number of the facial points employed, and due to differences in head-pose and facial expression. Most regression-based approaches for learning such mappings focus on modeling correlations only between the inputs (i.e., the facial points in a non-frontal pose) and the outputs (i.e., the facial points in the frontal pose), but not within the inputs and the outputs of the model. This makes these models prone to errors due to noise and outliers in test data, often resulting in anatomically impossible facial configurations formed by their predictions. To address this, we propose Shape-constrained Gaussian Process (SC-GP) regression for facial-point-based head-pose normalization. Specifically, a deformable face-shape model is used to learn a face-shape prior, which is placed on both the input and the output of GP regression in order to constrain the model predictions to anatomically feasible facial configurations. Our extensive experiments on both synthetic and real image data show that the proposed approach generalizes well across poses and handles successfully noise and outliers in test data. In addition, the proposed model outperforms previously proposed approaches to facial-point-based head-pose normalization.

AB - Given the facial points extracted from an image of a face in an arbitrary pose, the goal of facial-point-based headpose normalization is to obtain the corresponding facial points in a predefined pose (e.g., frontal). This involves inference of complex and high-dimensional mappings due to the large number of the facial points employed, and due to differences in head-pose and facial expression. Most regression-based approaches for learning such mappings focus on modeling correlations only between the inputs (i.e., the facial points in a non-frontal pose) and the outputs (i.e., the facial points in the frontal pose), but not within the inputs and the outputs of the model. This makes these models prone to errors due to noise and outliers in test data, often resulting in anatomically impossible facial configurations formed by their predictions. To address this, we propose Shape-constrained Gaussian Process (SC-GP) regression for facial-point-based head-pose normalization. Specifically, a deformable face-shape model is used to learn a face-shape prior, which is placed on both the input and the output of GP regression in order to constrain the model predictions to anatomically feasible facial configurations. Our extensive experiments on both synthetic and real image data show that the proposed approach generalizes well across poses and handles successfully noise and outliers in test data. In addition, the proposed model outperforms previously proposed approaches to facial-point-based head-pose normalization.

KW - METIS-285022

KW - IR-79458

KW - EWI-21316

KW - HMI-MI: MULTIMODAL INTERACTIONS

KW - EC Grant Agreement nr.: ERC/203143

U2 - 10.1109/ICCV.2011.6126407

DO - 10.1109/ICCV.2011.6126407

M3 - Conference contribution

SN - 978-1-4577-1101-5

SP - 1495

EP - 1502

BT - IEEE International Conference on Computer Vision, ICCV 2011

PB - IEEE Computer Society

CY - USA

ER -

Rudovic O, Pantic M. Shape-constrained Gaussian Process Regression for Facial-point-based Head-pose Normalization. In IEEE International Conference on Computer Vision, ICCV 2011. USA: IEEE Computer Society. 2011. p. 1495-1502 https://doi.org/10.1109/ICCV.2011.6126407