On dynamic scene geometry for view-invariant action matching

Anwaar Ul-Haq, Iqbal Gondal, Mohammad Murshed

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

    13 Citations (Scopus)

    Abstract

    Variation in viewpoints poses significant challenges to action recognition. One popular way of encoding view invariant-action representation is based on the exploitation of epipolar geometry between different views of the same action. Majority of representative work considers detection of landmark points and their tracking by assuming that motion trajectories for all landmark points on human body are available throughout the course of an action. Unfortunately, due to occlusion and noise, detection and tracking of these landmarks is not always robust. To facilitate it, some of the work assumes that such trajectories are manually marked which is a clear drawback and lacks automation introduced by computer vision. In this paper, we address this problem by proposing view invariant action matching score based on epipolar geometry between actor silhouettes, without tracking and explicit point correspondences. In addition, we explore multi-body epipolar constraint which facilitates to work on original action volumes without any pre-processing. We show that multibody fundamental matrix captures the geometry of dynamic
    action scenes and helps devising an action matching score across different views without any prior segmentation of actors. Extensive experimentation on challenging view invariant
    action datasets shows that our approach not only removes long standing assumptions but also achieves significant improvement in recognition accuracy and retrieval.
    Original languageEnglish
    Title of host publicationThe 24th IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2011)
    EditorsPedro Felzenszwalb, David Forsyth, Pascal Fua
    Place of PublicationLos Alamitos CA USA
    PublisherIEEE, Institute of Electrical and Electronics Engineers
    Pages3305-3312
    Number of pages8
    ISBN (Print)9781457703942
    DOIs
    Publication statusPublished - 2011
    EventIEEE Conference on Computer Vision and Pattern Recognition 2011 - Crowne Plaza, Colorado Springs, United States of America
    Duration: 20 Jun 201125 Jun 2011
    http://ieeexplore.ieee.org/xpl/mostRecentIssue.jsp?punumber=5968010 (IEEE Conference Proceedings)

    Conference

    ConferenceIEEE Conference on Computer Vision and Pattern Recognition 2011
    Abbreviated titleCVPR 2011
    CountryUnited States of America
    CityColorado Springs
    Period20/06/1125/06/11
    Internet address

    Cite this

    Ul-Haq, A., Gondal, I., & Murshed, M. (2011). On dynamic scene geometry for view-invariant action matching. In P. Felzenszwalb, D. Forsyth, & P. Fua (Eds.), The 24th IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2011) (pp. 3305-3312). Los Alamitos CA USA: IEEE, Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/CVPR.2011.5995690
    Ul-Haq, Anwaar ; Gondal, Iqbal ; Murshed, Mohammad. / On dynamic scene geometry for view-invariant action matching. The 24th IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2011). editor / Pedro Felzenszwalb ; David Forsyth ; Pascal Fua. Los Alamitos CA USA : IEEE, Institute of Electrical and Electronics Engineers, 2011. pp. 3305-3312
    @inproceedings{df4f1b413ae6421d97f385dc3c2d0c0d,
    title = "On dynamic scene geometry for view-invariant action matching",
    abstract = "Variation in viewpoints poses significant challenges to action recognition. One popular way of encoding view invariant-action representation is based on the exploitation of epipolar geometry between different views of the same action. Majority of representative work considers detection of landmark points and their tracking by assuming that motion trajectories for all landmark points on human body are available throughout the course of an action. Unfortunately, due to occlusion and noise, detection and tracking of these landmarks is not always robust. To facilitate it, some of the work assumes that such trajectories are manually marked which is a clear drawback and lacks automation introduced by computer vision. In this paper, we address this problem by proposing view invariant action matching score based on epipolar geometry between actor silhouettes, without tracking and explicit point correspondences. In addition, we explore multi-body epipolar constraint which facilitates to work on original action volumes without any pre-processing. We show that multibody fundamental matrix captures the geometry of dynamicaction scenes and helps devising an action matching score across different views without any prior segmentation of actors. Extensive experimentation on challenging view invariantaction datasets shows that our approach not only removes long standing assumptions but also achieves significant improvement in recognition accuracy and retrieval.",
    author = "Anwaar Ul-Haq and Iqbal Gondal and Mohammad Murshed",
    year = "2011",
    doi = "10.1109/CVPR.2011.5995690",
    language = "English",
    isbn = "9781457703942",
    pages = "3305--3312",
    editor = "Pedro Felzenszwalb and David Forsyth and Pascal Fua",
    booktitle = "The 24th IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2011)",
    publisher = "IEEE, Institute of Electrical and Electronics Engineers",
    address = "United States of America",

    }

    Ul-Haq, A, Gondal, I & Murshed, M 2011, On dynamic scene geometry for view-invariant action matching. in P Felzenszwalb, D Forsyth & P Fua (eds), The 24th IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2011). IEEE, Institute of Electrical and Electronics Engineers, Los Alamitos CA USA, pp. 3305-3312, IEEE Conference on Computer Vision and Pattern Recognition 2011, Colorado Springs, United States of America, 20/06/11. https://doi.org/10.1109/CVPR.2011.5995690

    On dynamic scene geometry for view-invariant action matching. / Ul-Haq, Anwaar; Gondal, Iqbal; Murshed, Mohammad.

    The 24th IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2011). ed. / Pedro Felzenszwalb; David Forsyth; Pascal Fua. Los Alamitos CA USA : IEEE, Institute of Electrical and Electronics Engineers, 2011. p. 3305-3312.

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

    TY - GEN

    T1 - On dynamic scene geometry for view-invariant action matching

    AU - Ul-Haq, Anwaar

    AU - Gondal, Iqbal

    AU - Murshed, Mohammad

    PY - 2011

    Y1 - 2011

    N2 - Variation in viewpoints poses significant challenges to action recognition. One popular way of encoding view invariant-action representation is based on the exploitation of epipolar geometry between different views of the same action. Majority of representative work considers detection of landmark points and their tracking by assuming that motion trajectories for all landmark points on human body are available throughout the course of an action. Unfortunately, due to occlusion and noise, detection and tracking of these landmarks is not always robust. To facilitate it, some of the work assumes that such trajectories are manually marked which is a clear drawback and lacks automation introduced by computer vision. In this paper, we address this problem by proposing view invariant action matching score based on epipolar geometry between actor silhouettes, without tracking and explicit point correspondences. In addition, we explore multi-body epipolar constraint which facilitates to work on original action volumes without any pre-processing. We show that multibody fundamental matrix captures the geometry of dynamicaction scenes and helps devising an action matching score across different views without any prior segmentation of actors. Extensive experimentation on challenging view invariantaction datasets shows that our approach not only removes long standing assumptions but also achieves significant improvement in recognition accuracy and retrieval.

    AB - Variation in viewpoints poses significant challenges to action recognition. One popular way of encoding view invariant-action representation is based on the exploitation of epipolar geometry between different views of the same action. Majority of representative work considers detection of landmark points and their tracking by assuming that motion trajectories for all landmark points on human body are available throughout the course of an action. Unfortunately, due to occlusion and noise, detection and tracking of these landmarks is not always robust. To facilitate it, some of the work assumes that such trajectories are manually marked which is a clear drawback and lacks automation introduced by computer vision. In this paper, we address this problem by proposing view invariant action matching score based on epipolar geometry between actor silhouettes, without tracking and explicit point correspondences. In addition, we explore multi-body epipolar constraint which facilitates to work on original action volumes without any pre-processing. We show that multibody fundamental matrix captures the geometry of dynamicaction scenes and helps devising an action matching score across different views without any prior segmentation of actors. Extensive experimentation on challenging view invariantaction datasets shows that our approach not only removes long standing assumptions but also achieves significant improvement in recognition accuracy and retrieval.

    U2 - 10.1109/CVPR.2011.5995690

    DO - 10.1109/CVPR.2011.5995690

    M3 - Conference Paper

    SN - 9781457703942

    SP - 3305

    EP - 3312

    BT - The 24th IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2011)

    A2 - Felzenszwalb, Pedro

    A2 - Forsyth, David

    A2 - Fua, Pascal

    PB - IEEE, Institute of Electrical and Electronics Engineers

    CY - Los Alamitos CA USA

    ER -

    Ul-Haq A, Gondal I, Murshed M. On dynamic scene geometry for view-invariant action matching. In Felzenszwalb P, Forsyth D, Fua P, editors, The 24th IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2011). Los Alamitos CA USA: IEEE, Institute of Electrical and Electronics Engineers. 2011. p. 3305-3312 https://doi.org/10.1109/CVPR.2011.5995690