Improving head and body pose estimation through semi-supervised manifold alignment

Alexandre Heili, Jagannadan Varadarajan, Bernard Ghanem, Narendra Ahuja, Jean Marc Odobez

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

Abstract

In this paper, we explore the use of a semi-supervised manifold alignment method for domain adaptation in the context of human body and head pose estimation in videos. We build upon an existing state-of-the-art system that leverages on external labelled datasets for the body and head features, and on the unlabelled test data with weak velocity labels to do a coupled estimation of the body and head pose. While this previous approach showed promising results, the learning of the underlying manifold structure of the features in the train and target data and the need to align them were not explored despite the fact that the pose features between two datasets may vary according to the scene, e.g. due to different camera point of view or perspective. In this paper, we propose to use a semi-supervised manifold alignment method to bring the train and target samples closer within the resulting embedded space. To this end, we consider an adaptation set from the target data and rely on (weak) labels, given for example by the velocity direction whenever they are reliable. These labels, along with the training labels are used to bias the manifold distance within each manifold and to establish correspondences for alignment.

Original languageEnglish (US)
Title of host publication2014 IEEE International Conference on Image Processing, ICIP 2014
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages1912-1916
Number of pages5
ISBN (Electronic)9781479957514
DOIs
StatePublished - Jan 28 2014

Publication series

Name2014 IEEE International Conference on Image Processing, ICIP 2014

Fingerprint

Labels
Cameras

Keywords

  • domain adaptation
  • head and body pose
  • manifold
  • semi-supervised
  • surveillance
  • weak labels

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition

Cite this

Heili, A., Varadarajan, J., Ghanem, B., Ahuja, N., & Odobez, J. M. (2014). Improving head and body pose estimation through semi-supervised manifold alignment. In 2014 IEEE International Conference on Image Processing, ICIP 2014 (pp. 1912-1916). [7025383] (2014 IEEE International Conference on Image Processing, ICIP 2014). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICIP.2014.7025383

Improving head and body pose estimation through semi-supervised manifold alignment. / Heili, Alexandre; Varadarajan, Jagannadan; Ghanem, Bernard; Ahuja, Narendra; Odobez, Jean Marc.

2014 IEEE International Conference on Image Processing, ICIP 2014. Institute of Electrical and Electronics Engineers Inc., 2014. p. 1912-1916 7025383 (2014 IEEE International Conference on Image Processing, ICIP 2014).

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

Heili, A, Varadarajan, J, Ghanem, B, Ahuja, N & Odobez, JM 2014, Improving head and body pose estimation through semi-supervised manifold alignment. in 2014 IEEE International Conference on Image Processing, ICIP 2014., 7025383, 2014 IEEE International Conference on Image Processing, ICIP 2014, Institute of Electrical and Electronics Engineers Inc., pp. 1912-1916. https://doi.org/10.1109/ICIP.2014.7025383
Heili A, Varadarajan J, Ghanem B, Ahuja N, Odobez JM. Improving head and body pose estimation through semi-supervised manifold alignment. In 2014 IEEE International Conference on Image Processing, ICIP 2014. Institute of Electrical and Electronics Engineers Inc. 2014. p. 1912-1916. 7025383. (2014 IEEE International Conference on Image Processing, ICIP 2014). https://doi.org/10.1109/ICIP.2014.7025383
Heili, Alexandre ; Varadarajan, Jagannadan ; Ghanem, Bernard ; Ahuja, Narendra ; Odobez, Jean Marc. / Improving head and body pose estimation through semi-supervised manifold alignment. 2014 IEEE International Conference on Image Processing, ICIP 2014. Institute of Electrical and Electronics Engineers Inc., 2014. pp. 1912-1916 (2014 IEEE International Conference on Image Processing, ICIP 2014).
@inproceedings{fd5513ea5ee944c2b00a78f5f57ac8c5,
title = "Improving head and body pose estimation through semi-supervised manifold alignment",
abstract = "In this paper, we explore the use of a semi-supervised manifold alignment method for domain adaptation in the context of human body and head pose estimation in videos. We build upon an existing state-of-the-art system that leverages on external labelled datasets for the body and head features, and on the unlabelled test data with weak velocity labels to do a coupled estimation of the body and head pose. While this previous approach showed promising results, the learning of the underlying manifold structure of the features in the train and target data and the need to align them were not explored despite the fact that the pose features between two datasets may vary according to the scene, e.g. due to different camera point of view or perspective. In this paper, we propose to use a semi-supervised manifold alignment method to bring the train and target samples closer within the resulting embedded space. To this end, we consider an adaptation set from the target data and rely on (weak) labels, given for example by the velocity direction whenever they are reliable. These labels, along with the training labels are used to bias the manifold distance within each manifold and to establish correspondences for alignment.",
keywords = "domain adaptation, head and body pose, manifold, semi-supervised, surveillance, weak labels",
author = "Alexandre Heili and Jagannadan Varadarajan and Bernard Ghanem and Narendra Ahuja and Odobez, {Jean Marc}",
year = "2014",
month = "1",
day = "28",
doi = "10.1109/ICIP.2014.7025383",
language = "English (US)",
series = "2014 IEEE International Conference on Image Processing, ICIP 2014",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "1912--1916",
booktitle = "2014 IEEE International Conference on Image Processing, ICIP 2014",
address = "United States",

}

TY - GEN

T1 - Improving head and body pose estimation through semi-supervised manifold alignment

AU - Heili, Alexandre

AU - Varadarajan, Jagannadan

AU - Ghanem, Bernard

AU - Ahuja, Narendra

AU - Odobez, Jean Marc

PY - 2014/1/28

Y1 - 2014/1/28

N2 - In this paper, we explore the use of a semi-supervised manifold alignment method for domain adaptation in the context of human body and head pose estimation in videos. We build upon an existing state-of-the-art system that leverages on external labelled datasets for the body and head features, and on the unlabelled test data with weak velocity labels to do a coupled estimation of the body and head pose. While this previous approach showed promising results, the learning of the underlying manifold structure of the features in the train and target data and the need to align them were not explored despite the fact that the pose features between two datasets may vary according to the scene, e.g. due to different camera point of view or perspective. In this paper, we propose to use a semi-supervised manifold alignment method to bring the train and target samples closer within the resulting embedded space. To this end, we consider an adaptation set from the target data and rely on (weak) labels, given for example by the velocity direction whenever they are reliable. These labels, along with the training labels are used to bias the manifold distance within each manifold and to establish correspondences for alignment.

AB - In this paper, we explore the use of a semi-supervised manifold alignment method for domain adaptation in the context of human body and head pose estimation in videos. We build upon an existing state-of-the-art system that leverages on external labelled datasets for the body and head features, and on the unlabelled test data with weak velocity labels to do a coupled estimation of the body and head pose. While this previous approach showed promising results, the learning of the underlying manifold structure of the features in the train and target data and the need to align them were not explored despite the fact that the pose features between two datasets may vary according to the scene, e.g. due to different camera point of view or perspective. In this paper, we propose to use a semi-supervised manifold alignment method to bring the train and target samples closer within the resulting embedded space. To this end, we consider an adaptation set from the target data and rely on (weak) labels, given for example by the velocity direction whenever they are reliable. These labels, along with the training labels are used to bias the manifold distance within each manifold and to establish correspondences for alignment.

KW - domain adaptation

KW - head and body pose

KW - manifold

KW - semi-supervised

KW - surveillance

KW - weak labels

UR - http://www.scopus.com/inward/record.url?scp=84956613534&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84956613534&partnerID=8YFLogxK

U2 - 10.1109/ICIP.2014.7025383

DO - 10.1109/ICIP.2014.7025383

M3 - Conference contribution

AN - SCOPUS:84956613534

T3 - 2014 IEEE International Conference on Image Processing, ICIP 2014

SP - 1912

EP - 1916

BT - 2014 IEEE International Conference on Image Processing, ICIP 2014

PB - Institute of Electrical and Electronics Engineers Inc.

ER -