Dominant Codewords Selection with Topic Model for Action Recognition

Hirokatsu Kataoka, Kenji Iwata, Yutaka Satoh, Masaki Hayashi, Yoshimitsu Aoki, Slobodan Ilic

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

Abstract

In this paper, we propose a framework for recognizing human activities that uses only in-topic dominant codewords and a mixture of intertopic vectors. Latent Dirichlet allocation (LDA) is used to develop approximations of human motion primitives, these are mid-level representations, and they adaptively integrate dominant vectors when classifying human activities. In LDA topic modeling, action videos (documents) are represented by a bag-of-words (input from a dictionary), and these are based on improved dense trajectories ([18]). The output topics correspond to human motion primitives, such as finger moving or subtle leg motion. We eliminate the impurities, such as missed tracking or changing light conditions, in each motion primitive. The assembled vector of motion primitives is an improved representation of the action. We demonstrate our method on four different datasets.

Original languageEnglish
Title of host publicationProceedings - 29th IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2016
PublisherIEEE Computer Society
Pages770-777
Number of pages8
ISBN (Electronic)9781467388504
DOIs
Publication statusPublished - 2016 Dec 16
Event29th IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2016 - Las Vegas, United States
Duration: 2016 Jun 262016 Jul 1

Other

Other29th IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2016
CountryUnited States
CityLas Vegas
Period16/6/2616/7/1

Fingerprint

Glossaries
Trajectories
Impurities

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Electrical and Electronic Engineering

Cite this

Kataoka, H., Iwata, K., Satoh, Y., Hayashi, M., Aoki, Y., & Ilic, S. (2016). Dominant Codewords Selection with Topic Model for Action Recognition. In Proceedings - 29th IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2016 (pp. 770-777). [7789591] IEEE Computer Society. https://doi.org/10.1109/CVPRW.2016.101

Dominant Codewords Selection with Topic Model for Action Recognition. / Kataoka, Hirokatsu; Iwata, Kenji; Satoh, Yutaka; Hayashi, Masaki; Aoki, Yoshimitsu; Ilic, Slobodan.

Proceedings - 29th IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2016. IEEE Computer Society, 2016. p. 770-777 7789591.

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

Kataoka, H, Iwata, K, Satoh, Y, Hayashi, M, Aoki, Y & Ilic, S 2016, Dominant Codewords Selection with Topic Model for Action Recognition. in Proceedings - 29th IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2016., 7789591, IEEE Computer Society, pp. 770-777, 29th IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2016, Las Vegas, United States, 16/6/26. https://doi.org/10.1109/CVPRW.2016.101
Kataoka H, Iwata K, Satoh Y, Hayashi M, Aoki Y, Ilic S. Dominant Codewords Selection with Topic Model for Action Recognition. In Proceedings - 29th IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2016. IEEE Computer Society. 2016. p. 770-777. 7789591 https://doi.org/10.1109/CVPRW.2016.101
Kataoka, Hirokatsu ; Iwata, Kenji ; Satoh, Yutaka ; Hayashi, Masaki ; Aoki, Yoshimitsu ; Ilic, Slobodan. / Dominant Codewords Selection with Topic Model for Action Recognition. Proceedings - 29th IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2016. IEEE Computer Society, 2016. pp. 770-777
@inproceedings{02d3c9c0d5e24632b4af5305946a8272,
title = "Dominant Codewords Selection with Topic Model for Action Recognition",
abstract = "In this paper, we propose a framework for recognizing human activities that uses only in-topic dominant codewords and a mixture of intertopic vectors. Latent Dirichlet allocation (LDA) is used to develop approximations of human motion primitives, these are mid-level representations, and they adaptively integrate dominant vectors when classifying human activities. In LDA topic modeling, action videos (documents) are represented by a bag-of-words (input from a dictionary), and these are based on improved dense trajectories ([18]). The output topics correspond to human motion primitives, such as finger moving or subtle leg motion. We eliminate the impurities, such as missed tracking or changing light conditions, in each motion primitive. The assembled vector of motion primitives is an improved representation of the action. We demonstrate our method on four different datasets.",
author = "Hirokatsu Kataoka and Kenji Iwata and Yutaka Satoh and Masaki Hayashi and Yoshimitsu Aoki and Slobodan Ilic",
year = "2016",
month = "12",
day = "16",
doi = "10.1109/CVPRW.2016.101",
language = "English",
pages = "770--777",
booktitle = "Proceedings - 29th IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2016",
publisher = "IEEE Computer Society",

}

TY - GEN

T1 - Dominant Codewords Selection with Topic Model for Action Recognition

AU - Kataoka, Hirokatsu

AU - Iwata, Kenji

AU - Satoh, Yutaka

AU - Hayashi, Masaki

AU - Aoki, Yoshimitsu

AU - Ilic, Slobodan

PY - 2016/12/16

Y1 - 2016/12/16

N2 - In this paper, we propose a framework for recognizing human activities that uses only in-topic dominant codewords and a mixture of intertopic vectors. Latent Dirichlet allocation (LDA) is used to develop approximations of human motion primitives, these are mid-level representations, and they adaptively integrate dominant vectors when classifying human activities. In LDA topic modeling, action videos (documents) are represented by a bag-of-words (input from a dictionary), and these are based on improved dense trajectories ([18]). The output topics correspond to human motion primitives, such as finger moving or subtle leg motion. We eliminate the impurities, such as missed tracking or changing light conditions, in each motion primitive. The assembled vector of motion primitives is an improved representation of the action. We demonstrate our method on four different datasets.

AB - In this paper, we propose a framework for recognizing human activities that uses only in-topic dominant codewords and a mixture of intertopic vectors. Latent Dirichlet allocation (LDA) is used to develop approximations of human motion primitives, these are mid-level representations, and they adaptively integrate dominant vectors when classifying human activities. In LDA topic modeling, action videos (documents) are represented by a bag-of-words (input from a dictionary), and these are based on improved dense trajectories ([18]). The output topics correspond to human motion primitives, such as finger moving or subtle leg motion. We eliminate the impurities, such as missed tracking or changing light conditions, in each motion primitive. The assembled vector of motion primitives is an improved representation of the action. We demonstrate our method on four different datasets.

UR - http://www.scopus.com/inward/record.url?scp=85010216531&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85010216531&partnerID=8YFLogxK

U2 - 10.1109/CVPRW.2016.101

DO - 10.1109/CVPRW.2016.101

M3 - Conference contribution

AN - SCOPUS:85010216531

SP - 770

EP - 777

BT - Proceedings - 29th IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2016

PB - IEEE Computer Society

ER -