Multi-Conditional Latent Variable Model for Joint Facial Action Unit Detection

Stefanos Eleftheriadis, Ognjen Rudovic, Maja Pantic

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

40 Citations (Scopus)
1 Downloads (Pure)

Abstract

We propose a novel multi-conditional latent variable model for simultaneous facial feature fusion and detection of facial action units. In our approach we exploit the structure-discovery capabilities of generative models such as Gaussian processes, and the discriminative power of classifiers such as logistic function. This leads to superior performance compared to existing classifiers for the target task that exploit either the discriminative or generative property, but not both. The model learning is performed via an efficient, newly proposed Bayesian learning strategy based on Monte Carlo sampling. Consequently, the learned model is robust to data overfitting, regardless of the number of both input features and jointly estimated facial action units. Extensive qualitative and quantitative experimental evaluations are performed on three publicly available datasets (CK+, Shoulder-pain and DISFA). We show that the proposed model outperforms the state-of-the-art methods for the target task on (i) feature fusion, and (ii) multiple facial action unit detection.
Original languageEnglish
Title of host publicationProceedings of the IEEE International Conference on Computer Vision (ICCV 2015)
Place of PublicationUSA
PublisherIEEE
Pages3792-3800
Number of pages9
DOIs
Publication statusPublished - Dec 2015
EventIEEE International Conference on Computer Vision 2015 - Convention Center in Santiago, Santiago, Chile
Duration: 7 Dec 201513 Dec 2015
http://pamitc.org/iccv15/

Publication series

Name
PublisherIEEE

Workshop

WorkshopIEEE International Conference on Computer Vision 2015
Abbreviated titleICCV 2015
CountryChile
CitySantiago
Period7/12/1513/12/15
Internet address

Fingerprint

Classifiers
Fusion reactions
Logistics
Sampling

Keywords

  • HMI-HF: Human Factors
  • EC Grant Agreement nr.: FP7/611153
  • IR-99579
  • METIS-316050
  • EWI-26841

Cite this

Eleftheriadis, S., Rudovic, O., & Pantic, M. (2015). Multi-Conditional Latent Variable Model for Joint Facial Action Unit Detection. In Proceedings of the IEEE International Conference on Computer Vision (ICCV 2015) (pp. 3792-3800). USA: IEEE. https://doi.org/10.1109/ICCV.2015.432
Eleftheriadis, Stefanos ; Rudovic, Ognjen ; Pantic, Maja. / Multi-Conditional Latent Variable Model for Joint Facial Action Unit Detection. Proceedings of the IEEE International Conference on Computer Vision (ICCV 2015). USA : IEEE, 2015. pp. 3792-3800
@inproceedings{791742fdd086448d97bd257caa496c3d,
title = "Multi-Conditional Latent Variable Model for Joint Facial Action Unit Detection",
abstract = "We propose a novel multi-conditional latent variable model for simultaneous facial feature fusion and detection of facial action units. In our approach we exploit the structure-discovery capabilities of generative models such as Gaussian processes, and the discriminative power of classifiers such as logistic function. This leads to superior performance compared to existing classifiers for the target task that exploit either the discriminative or generative property, but not both. The model learning is performed via an efficient, newly proposed Bayesian learning strategy based on Monte Carlo sampling. Consequently, the learned model is robust to data overfitting, regardless of the number of both input features and jointly estimated facial action units. Extensive qualitative and quantitative experimental evaluations are performed on three publicly available datasets (CK+, Shoulder-pain and DISFA). We show that the proposed model outperforms the state-of-the-art methods for the target task on (i) feature fusion, and (ii) multiple facial action unit detection.",
keywords = "HMI-HF: Human Factors, EC Grant Agreement nr.: FP7/611153, IR-99579, METIS-316050, EWI-26841",
author = "Stefanos Eleftheriadis and Ognjen Rudovic and Maja Pantic",
note = "eemcs-eprint-26841",
year = "2015",
month = "12",
doi = "10.1109/ICCV.2015.432",
language = "English",
publisher = "IEEE",
pages = "3792--3800",
booktitle = "Proceedings of the IEEE International Conference on Computer Vision (ICCV 2015)",
address = "United States",

}

Eleftheriadis, S, Rudovic, O & Pantic, M 2015, Multi-Conditional Latent Variable Model for Joint Facial Action Unit Detection. in Proceedings of the IEEE International Conference on Computer Vision (ICCV 2015). IEEE, USA, pp. 3792-3800, IEEE International Conference on Computer Vision 2015, Santiago, Chile, 7/12/15. https://doi.org/10.1109/ICCV.2015.432

Multi-Conditional Latent Variable Model for Joint Facial Action Unit Detection. / Eleftheriadis, Stefanos; Rudovic, Ognjen; Pantic, Maja.

Proceedings of the IEEE International Conference on Computer Vision (ICCV 2015). USA : IEEE, 2015. p. 3792-3800.

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

TY - GEN

T1 - Multi-Conditional Latent Variable Model for Joint Facial Action Unit Detection

AU - Eleftheriadis, Stefanos

AU - Rudovic, Ognjen

AU - Pantic, Maja

N1 - eemcs-eprint-26841

PY - 2015/12

Y1 - 2015/12

N2 - We propose a novel multi-conditional latent variable model for simultaneous facial feature fusion and detection of facial action units. In our approach we exploit the structure-discovery capabilities of generative models such as Gaussian processes, and the discriminative power of classifiers such as logistic function. This leads to superior performance compared to existing classifiers for the target task that exploit either the discriminative or generative property, but not both. The model learning is performed via an efficient, newly proposed Bayesian learning strategy based on Monte Carlo sampling. Consequently, the learned model is robust to data overfitting, regardless of the number of both input features and jointly estimated facial action units. Extensive qualitative and quantitative experimental evaluations are performed on three publicly available datasets (CK+, Shoulder-pain and DISFA). We show that the proposed model outperforms the state-of-the-art methods for the target task on (i) feature fusion, and (ii) multiple facial action unit detection.

AB - We propose a novel multi-conditional latent variable model for simultaneous facial feature fusion and detection of facial action units. In our approach we exploit the structure-discovery capabilities of generative models such as Gaussian processes, and the discriminative power of classifiers such as logistic function. This leads to superior performance compared to existing classifiers for the target task that exploit either the discriminative or generative property, but not both. The model learning is performed via an efficient, newly proposed Bayesian learning strategy based on Monte Carlo sampling. Consequently, the learned model is robust to data overfitting, regardless of the number of both input features and jointly estimated facial action units. Extensive qualitative and quantitative experimental evaluations are performed on three publicly available datasets (CK+, Shoulder-pain and DISFA). We show that the proposed model outperforms the state-of-the-art methods for the target task on (i) feature fusion, and (ii) multiple facial action unit detection.

KW - HMI-HF: Human Factors

KW - EC Grant Agreement nr.: FP7/611153

KW - IR-99579

KW - METIS-316050

KW - EWI-26841

U2 - 10.1109/ICCV.2015.432

DO - 10.1109/ICCV.2015.432

M3 - Conference contribution

SP - 3792

EP - 3800

BT - Proceedings of the IEEE International Conference on Computer Vision (ICCV 2015)

PB - IEEE

CY - USA

ER -

Eleftheriadis S, Rudovic O, Pantic M. Multi-Conditional Latent Variable Model for Joint Facial Action Unit Detection. In Proceedings of the IEEE International Conference on Computer Vision (ICCV 2015). USA: IEEE. 2015. p. 3792-3800 https://doi.org/10.1109/ICCV.2015.432