Ordered Trajectories for Large Scale Human Action Recognition

Ramana ORUGANTI, Roland GOECKE

Research output: A Conference proceeding or a Chapter in BookConference contribution

28 Citations (Scopus)

Abstract

Recently, a video representation based on dense trajectories has been shown to outperform other human action recognition methods on several benchmark datasets. In dense trajectories, points are sampled at uniform intervals in space and time and then tracked using a dense optical flow field. The uniform sampling does not discriminate objects of interest from the background or other objects. Consequently, a lot of information is accumulated, which actually may not be useful. Sometimes, this unwanted information may bias the learning process if its content is much larger than the information of the principal object(s) of interest. This can especially escalate when more and more data is accumulated due to an increase in the number of action classes or the computation of dense trajectories at different scales in space and time, as in the Spatio-Temporal Pyramidal approach. In contrast, we propose a technique that selects only a few dense trajectories and then generates a new set of trajectories termed 'ordered trajectories'. We evaluate our technique on the complex benchmark HMDB51, UCF50 and UCF101 datasets containing 50 or more action classes and observe improved performance in terms of recognition rates and removal of background clutter at a lower computational cost.
Original languageEnglish
Title of host publicationProceedings of the IEEE International Conference on Computer Vision (ICCV2013) Workshops
EditorsIvan Laptev, Massimo Piccardi, Mubarak Shah, Rahul Sukthankar
Place of PublicationPiscataway, USA
PublisherIEEE, Institute of Electrical and Electronics Engineers
Pages412-419
Number of pages8
ISBN (Electronic)9781479930227
DOIs
Publication statusPublished - 2 Dec 2013
EventIEEE International Conference on Computer Vision (ICCV2013) - Sydney, Sydney, Australia
Duration: 1 Dec 20138 Dec 2013

Conference

ConferenceIEEE International Conference on Computer Vision (ICCV2013)
CountryAustralia
CitySydney
Period1/12/138/12/13

Fingerprint

Trajectories
Optical flows
Flow fields
Sampling
Costs

Cite this

ORUGANTI, R., & GOECKE, R. (2013). Ordered Trajectories for Large Scale Human Action Recognition. In I. Laptev, M. Piccardi, M. Shah, & R. Sukthankar (Eds.), Proceedings of the IEEE International Conference on Computer Vision (ICCV2013) Workshops (pp. 412-419). Piscataway, USA: IEEE, Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/ICCVW.2013.61
ORUGANTI, Ramana ; GOECKE, Roland. / Ordered Trajectories for Large Scale Human Action Recognition. Proceedings of the IEEE International Conference on Computer Vision (ICCV2013) Workshops. editor / Ivan Laptev ; Massimo Piccardi ; Mubarak Shah ; Rahul Sukthankar. Piscataway, USA : IEEE, Institute of Electrical and Electronics Engineers, 2013. pp. 412-419
@inproceedings{d9b61d44ec3645408a6714446943b540,
title = "Ordered Trajectories for Large Scale Human Action Recognition",
abstract = "Recently, a video representation based on dense trajectories has been shown to outperform other human action recognition methods on several benchmark datasets. In dense trajectories, points are sampled at uniform intervals in space and time and then tracked using a dense optical flow field. The uniform sampling does not discriminate objects of interest from the background or other objects. Consequently, a lot of information is accumulated, which actually may not be useful. Sometimes, this unwanted information may bias the learning process if its content is much larger than the information of the principal object(s) of interest. This can especially escalate when more and more data is accumulated due to an increase in the number of action classes or the computation of dense trajectories at different scales in space and time, as in the Spatio-Temporal Pyramidal approach. In contrast, we propose a technique that selects only a few dense trajectories and then generates a new set of trajectories termed 'ordered trajectories'. We evaluate our technique on the complex benchmark HMDB51, UCF50 and UCF101 datasets containing 50 or more action classes and observe improved performance in terms of recognition rates and removal of background clutter at a lower computational cost.",
keywords = "Human action recognition, Large Scale, Ordered trajectories",
author = "Ramana ORUGANTI and Roland GOECKE",
year = "2013",
month = "12",
day = "2",
doi = "10.1109/ICCVW.2013.61",
language = "English",
pages = "412--419",
editor = "Ivan Laptev and Massimo Piccardi and Mubarak Shah and Rahul Sukthankar",
booktitle = "Proceedings of the IEEE International Conference on Computer Vision (ICCV2013) Workshops",
publisher = "IEEE, Institute of Electrical and Electronics Engineers",
address = "United States",

}

ORUGANTI, R & GOECKE, R 2013, Ordered Trajectories for Large Scale Human Action Recognition. in I Laptev, M Piccardi, M Shah & R Sukthankar (eds), Proceedings of the IEEE International Conference on Computer Vision (ICCV2013) Workshops. IEEE, Institute of Electrical and Electronics Engineers, Piscataway, USA, pp. 412-419, IEEE International Conference on Computer Vision (ICCV2013), Sydney, Australia, 1/12/13. https://doi.org/10.1109/ICCVW.2013.61

Ordered Trajectories for Large Scale Human Action Recognition. / ORUGANTI, Ramana; GOECKE, Roland.

Proceedings of the IEEE International Conference on Computer Vision (ICCV2013) Workshops. ed. / Ivan Laptev; Massimo Piccardi; Mubarak Shah; Rahul Sukthankar. Piscataway, USA : IEEE, Institute of Electrical and Electronics Engineers, 2013. p. 412-419.

Research output: A Conference proceeding or a Chapter in BookConference contribution

TY - GEN

T1 - Ordered Trajectories for Large Scale Human Action Recognition

AU - ORUGANTI, Ramana

AU - GOECKE, Roland

PY - 2013/12/2

Y1 - 2013/12/2

N2 - Recently, a video representation based on dense trajectories has been shown to outperform other human action recognition methods on several benchmark datasets. In dense trajectories, points are sampled at uniform intervals in space and time and then tracked using a dense optical flow field. The uniform sampling does not discriminate objects of interest from the background or other objects. Consequently, a lot of information is accumulated, which actually may not be useful. Sometimes, this unwanted information may bias the learning process if its content is much larger than the information of the principal object(s) of interest. This can especially escalate when more and more data is accumulated due to an increase in the number of action classes or the computation of dense trajectories at different scales in space and time, as in the Spatio-Temporal Pyramidal approach. In contrast, we propose a technique that selects only a few dense trajectories and then generates a new set of trajectories termed 'ordered trajectories'. We evaluate our technique on the complex benchmark HMDB51, UCF50 and UCF101 datasets containing 50 or more action classes and observe improved performance in terms of recognition rates and removal of background clutter at a lower computational cost.

AB - Recently, a video representation based on dense trajectories has been shown to outperform other human action recognition methods on several benchmark datasets. In dense trajectories, points are sampled at uniform intervals in space and time and then tracked using a dense optical flow field. The uniform sampling does not discriminate objects of interest from the background or other objects. Consequently, a lot of information is accumulated, which actually may not be useful. Sometimes, this unwanted information may bias the learning process if its content is much larger than the information of the principal object(s) of interest. This can especially escalate when more and more data is accumulated due to an increase in the number of action classes or the computation of dense trajectories at different scales in space and time, as in the Spatio-Temporal Pyramidal approach. In contrast, we propose a technique that selects only a few dense trajectories and then generates a new set of trajectories termed 'ordered trajectories'. We evaluate our technique on the complex benchmark HMDB51, UCF50 and UCF101 datasets containing 50 or more action classes and observe improved performance in terms of recognition rates and removal of background clutter at a lower computational cost.

KW - Human action recognition

KW - Large Scale

KW - Ordered trajectories

U2 - 10.1109/ICCVW.2013.61

DO - 10.1109/ICCVW.2013.61

M3 - Conference contribution

SP - 412

EP - 419

BT - Proceedings of the IEEE International Conference on Computer Vision (ICCV2013) Workshops

A2 - Laptev, Ivan

A2 - Piccardi, Massimo

A2 - Shah, Mubarak

A2 - Sukthankar, Rahul

PB - IEEE, Institute of Electrical and Electronics Engineers

CY - Piscataway, USA

ER -

ORUGANTI R, GOECKE R. Ordered Trajectories for Large Scale Human Action Recognition. In Laptev I, Piccardi M, Shah M, Sukthankar R, editors, Proceedings of the IEEE International Conference on Computer Vision (ICCV2013) Workshops. Piscataway, USA: IEEE, Institute of Electrical and Electronics Engineers. 2013. p. 412-419 https://doi.org/10.1109/ICCVW.2013.61