Standard

Using Phase Instead of Optical Flow for Action Recognition. / Hommos, Omar; Pintea, Silvia L.; Mettes, Pascal S.M.; van Gemert, Jan C.

Computer Vision – ECCV 2018 Workshops, Proceedings. ed. / Laura Leal-Taixé; Stefan Roth. part VI. ed. Cham : Springer, 2019. p. 678-691 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11134 LNCS).

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

Harvard

Hommos, O, Pintea, SL, Mettes, PSM & van Gemert, JC 2019, Using Phase Instead of Optical Flow for Action Recognition. in L Leal-Taixé & S Roth (eds), Computer Vision – ECCV 2018 Workshops, Proceedings. part VI edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11134 LNCS, Springer, Cham, pp. 678-691, 15th European Conference on Computer Vision, ECCV 2018, Munich, Germany, 8/09/18. https://doi.org/10.1007/978-3-030-11024-6_51

APA

Hommos, O., Pintea, S. L., Mettes, P. S. M., & van Gemert, J. C. (2019). Using Phase Instead of Optical Flow for Action Recognition. In L. Leal-Taixé, & S. Roth (Eds.), Computer Vision – ECCV 2018 Workshops, Proceedings (part VI ed., pp. 678-691). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11134 LNCS). Cham: Springer. https://doi.org/10.1007/978-3-030-11024-6_51

Vancouver

Hommos O, Pintea SL, Mettes PSM, van Gemert JC. Using Phase Instead of Optical Flow for Action Recognition. In Leal-Taixé L, Roth S, editors, Computer Vision – ECCV 2018 Workshops, Proceedings. part VI ed. Cham: Springer. 2019. p. 678-691. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-11024-6_51

Author

Hommos, Omar ; Pintea, Silvia L. ; Mettes, Pascal S.M. ; van Gemert, Jan C. / Using Phase Instead of Optical Flow for Action Recognition. Computer Vision – ECCV 2018 Workshops, Proceedings. editor / Laura Leal-Taixé ; Stefan Roth. part VI. ed. Cham : Springer, 2019. pp. 678-691 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

BibTeX

@inproceedings{08d6fb46415f461291e3543f3c48a52b,
title = "Using Phase Instead of Optical Flow for Action Recognition",
abstract = "Currently, the most common motion representation for action recognition is optical flow. Optical flow is based on particle tracking which adheres to a Lagrangian perspective on dynamics. In contrast to the Lagrangian perspective, the Eulerian model of dynamics does not track, but describes local changes. For video, an Eulerian phase-based motion representation, using complex steerable filters, has been successfully employed recently for motion magnification and video frame interpolation. Inspired by these previous works, here, we proposes learning Eulerian motion representations in a deep architecture for action recognition. We learn filters in the complex domain in an end-to-end manner. We design these complex filters to resemble complex Gabor filters, typically employed for phase-information extraction. We propose a phase-information extraction module, based on these complex filters, that can be used in any network architecture for extracting Eulerian representations. We experimentally analyze the added value of Eulerian motion representations, as extracted by our proposed phase extraction module, and compare with existing motion representations based on optical flow, on the UCF101 dataset.",
keywords = "Action recognition, Eulerian motion representation, Motion representation, Phase derivatives",
author = "Omar Hommos and Pintea, {Silvia L.} and Mettes, {Pascal S.M.} and {van Gemert}, {Jan C.}",
year = "2019",
doi = "10.1007/978-3-030-11024-6_51",
language = "English",
isbn = "978-303011023-9",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer",
pages = "678--691",
editor = "Laura Leal-Taix{\'e} and Stefan Roth",
booktitle = "Computer Vision – ECCV 2018 Workshops, Proceedings",
edition = "part VI",

}

RIS

TY - GEN

T1 - Using Phase Instead of Optical Flow for Action Recognition

AU - Hommos, Omar

AU - Pintea, Silvia L.

AU - Mettes, Pascal S.M.

AU - van Gemert, Jan C.

PY - 2019

Y1 - 2019

N2 - Currently, the most common motion representation for action recognition is optical flow. Optical flow is based on particle tracking which adheres to a Lagrangian perspective on dynamics. In contrast to the Lagrangian perspective, the Eulerian model of dynamics does not track, but describes local changes. For video, an Eulerian phase-based motion representation, using complex steerable filters, has been successfully employed recently for motion magnification and video frame interpolation. Inspired by these previous works, here, we proposes learning Eulerian motion representations in a deep architecture for action recognition. We learn filters in the complex domain in an end-to-end manner. We design these complex filters to resemble complex Gabor filters, typically employed for phase-information extraction. We propose a phase-information extraction module, based on these complex filters, that can be used in any network architecture for extracting Eulerian representations. We experimentally analyze the added value of Eulerian motion representations, as extracted by our proposed phase extraction module, and compare with existing motion representations based on optical flow, on the UCF101 dataset.

AB - Currently, the most common motion representation for action recognition is optical flow. Optical flow is based on particle tracking which adheres to a Lagrangian perspective on dynamics. In contrast to the Lagrangian perspective, the Eulerian model of dynamics does not track, but describes local changes. For video, an Eulerian phase-based motion representation, using complex steerable filters, has been successfully employed recently for motion magnification and video frame interpolation. Inspired by these previous works, here, we proposes learning Eulerian motion representations in a deep architecture for action recognition. We learn filters in the complex domain in an end-to-end manner. We design these complex filters to resemble complex Gabor filters, typically employed for phase-information extraction. We propose a phase-information extraction module, based on these complex filters, that can be used in any network architecture for extracting Eulerian representations. We experimentally analyze the added value of Eulerian motion representations, as extracted by our proposed phase extraction module, and compare with existing motion representations based on optical flow, on the UCF101 dataset.

KW - Action recognition

KW - Eulerian motion representation

KW - Motion representation

KW - Phase derivatives

UR - http://www.scopus.com/inward/record.url?scp=85061720021&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-11024-6_51

DO - 10.1007/978-3-030-11024-6_51

M3 - Conference contribution

SN - 978-303011023-9

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 678

EP - 691

BT - Computer Vision – ECCV 2018 Workshops, Proceedings

A2 - Leal-Taixé, Laura

A2 - Roth, Stefan

PB - Springer

CY - Cham

ER -

ID: 51734703