Extended co-occurrence HOG with dense trajectories for fine-grained activity recognition

Hirokatsu Kataoka, Kiyoshi Hashimoto, Kenji Iwata, Yutaka Satoh, Nassir Navab, Slobodan Ilic, Yoshimitsu Aoki

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

9 Citations (Scopus)

Abstract

In this paper we propose a novel feature descriptor Extended Co-occurrence HOG (ECoHOG) and integrate it with dense point trajectories demonstrating its usefulness in fine grained activity recognition. This feature is inspired by original Co-occurrence HOG (CoHOG) that is based on histograms of occurrences of pairs of image gradients in the image. Instead relying only on pure histograms we introduce a sum of gradient magnitudes of co-occurring pairs of image gradients in the image. This results in giving the importance to the object boundaries and straightening the difference between the moving foreground and static background. We also couple ECoHOG with dense point trajectories extracted using optical flow from video sequences and demonstrate that they are extremely well suited for fine grained activity recognition. Using our feature we outperform state of the art methods in this task and provide extensive quantitative evaluation.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages336-349
Number of pages14
Volume9007
ISBN (Print)9783319168135
DOIs
Publication statusPublished - 2015
Event12th Asian Conference on Computer Vision, ACCV 2014 - Singapore, Singapore
Duration: 2014 Nov 12014 Nov 5

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9007
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other12th Asian Conference on Computer Vision, ACCV 2014
CountrySingapore
CitySingapore
Period14/11/114/11/5

Fingerprint

Activity Recognition
Trajectories
Trajectory
Straightening
Optical flows
Gradient
Histogram
Quantitative Evaluation
Optical Flow
Descriptors
Integrate
Demonstrate

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Kataoka, H., Hashimoto, K., Iwata, K., Satoh, Y., Navab, N., Ilic, S., & Aoki, Y. (2015). Extended co-occurrence HOG with dense trajectories for fine-grained activity recognition. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9007, pp. 336-349). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9007). Springer Verlag. https://doi.org/10.1007/978-3-319-16814-2_22

Extended co-occurrence HOG with dense trajectories for fine-grained activity recognition. / Kataoka, Hirokatsu; Hashimoto, Kiyoshi; Iwata, Kenji; Satoh, Yutaka; Navab, Nassir; Ilic, Slobodan; Aoki, Yoshimitsu.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9007 Springer Verlag, 2015. p. 336-349 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9007).

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

Kataoka, H, Hashimoto, K, Iwata, K, Satoh, Y, Navab, N, Ilic, S & Aoki, Y 2015, Extended co-occurrence HOG with dense trajectories for fine-grained activity recognition. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 9007, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9007, Springer Verlag, pp. 336-349, 12th Asian Conference on Computer Vision, ACCV 2014, Singapore, Singapore, 14/11/1. https://doi.org/10.1007/978-3-319-16814-2_22
Kataoka H, Hashimoto K, Iwata K, Satoh Y, Navab N, Ilic S et al. Extended co-occurrence HOG with dense trajectories for fine-grained activity recognition. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9007. Springer Verlag. 2015. p. 336-349. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-16814-2_22
Kataoka, Hirokatsu ; Hashimoto, Kiyoshi ; Iwata, Kenji ; Satoh, Yutaka ; Navab, Nassir ; Ilic, Slobodan ; Aoki, Yoshimitsu. / Extended co-occurrence HOG with dense trajectories for fine-grained activity recognition. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9007 Springer Verlag, 2015. pp. 336-349 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{625d795bf94744b793da1bfd4374d076,
title = "Extended co-occurrence HOG with dense trajectories for fine-grained activity recognition",
abstract = "In this paper we propose a novel feature descriptor Extended Co-occurrence HOG (ECoHOG) and integrate it with dense point trajectories demonstrating its usefulness in fine grained activity recognition. This feature is inspired by original Co-occurrence HOG (CoHOG) that is based on histograms of occurrences of pairs of image gradients in the image. Instead relying only on pure histograms we introduce a sum of gradient magnitudes of co-occurring pairs of image gradients in the image. This results in giving the importance to the object boundaries and straightening the difference between the moving foreground and static background. We also couple ECoHOG with dense point trajectories extracted using optical flow from video sequences and demonstrate that they are extremely well suited for fine grained activity recognition. Using our feature we outperform state of the art methods in this task and provide extensive quantitative evaluation.",
author = "Hirokatsu Kataoka and Kiyoshi Hashimoto and Kenji Iwata and Yutaka Satoh and Nassir Navab and Slobodan Ilic and Yoshimitsu Aoki",
year = "2015",
doi = "10.1007/978-3-319-16814-2_22",
language = "English",
isbn = "9783319168135",
volume = "9007",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "336--349",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Extended co-occurrence HOG with dense trajectories for fine-grained activity recognition

AU - Kataoka, Hirokatsu

AU - Hashimoto, Kiyoshi

AU - Iwata, Kenji

AU - Satoh, Yutaka

AU - Navab, Nassir

AU - Ilic, Slobodan

AU - Aoki, Yoshimitsu

PY - 2015

Y1 - 2015

N2 - In this paper we propose a novel feature descriptor Extended Co-occurrence HOG (ECoHOG) and integrate it with dense point trajectories demonstrating its usefulness in fine grained activity recognition. This feature is inspired by original Co-occurrence HOG (CoHOG) that is based on histograms of occurrences of pairs of image gradients in the image. Instead relying only on pure histograms we introduce a sum of gradient magnitudes of co-occurring pairs of image gradients in the image. This results in giving the importance to the object boundaries and straightening the difference between the moving foreground and static background. We also couple ECoHOG with dense point trajectories extracted using optical flow from video sequences and demonstrate that they are extremely well suited for fine grained activity recognition. Using our feature we outperform state of the art methods in this task and provide extensive quantitative evaluation.

AB - In this paper we propose a novel feature descriptor Extended Co-occurrence HOG (ECoHOG) and integrate it with dense point trajectories demonstrating its usefulness in fine grained activity recognition. This feature is inspired by original Co-occurrence HOG (CoHOG) that is based on histograms of occurrences of pairs of image gradients in the image. Instead relying only on pure histograms we introduce a sum of gradient magnitudes of co-occurring pairs of image gradients in the image. This results in giving the importance to the object boundaries and straightening the difference between the moving foreground and static background. We also couple ECoHOG with dense point trajectories extracted using optical flow from video sequences and demonstrate that they are extremely well suited for fine grained activity recognition. Using our feature we outperform state of the art methods in this task and provide extensive quantitative evaluation.

UR - http://www.scopus.com/inward/record.url?scp=84929630051&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84929630051&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-16814-2_22

DO - 10.1007/978-3-319-16814-2_22

M3 - Conference contribution

AN - SCOPUS:84929630051

SN - 9783319168135

VL - 9007

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

SP - 336

EP - 349

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

PB - Springer Verlag

ER -