Evaluation of vision-based human activity recognition in dense trajectory framework

Hirokatsu Kataoka, Yoshimitsu Aoki, Kenji Iwata, Yutaka Satoh

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

3 Citations (Scopus)

Abstract

Activity recognition has been an active research topic in computer vision. Recently, the most successful approaches use dense trajectories that extract a large number of trajectories and features on the trajectories into a codeword. In this paper, we evaluate various features in the framework of dense trajectories on several types of datasets. We implement 13 features in total by including five different types of descriptor, namely motion-, shape-, texture- trajectory- and co-occurrence-based feature descriptors. The experimental results show a relationship between feature descriptors and performance rate at each dataset. Different scenes of traffic, surgery, daily living and sports are used to analyze the feature characteristics. Moreover, we test how much the performance rate of concatenated vectors depends on the type, top-ranked in experiment and all 13 feature descriptors on fine-grained datasets. Feature evaluation is beneficial not only in the activity recognition problem, but also in other domains in spatio-temporal recognition.

Original languageEnglish
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PublisherSpringer Verlag
Pages634-646
Number of pages13
Volume9474
ISBN (Print)9783319278568
DOIs
Publication statusPublished - 2015
Event11th International Symposium on Advances in Visual Computing, ISVC 2015 - Las Vegas, United States
Duration: 2015 Dec 142015 Dec 16

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9474
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other11th International Symposium on Advances in Visual Computing, ISVC 2015
CountryUnited States
CityLas Vegas
Period15/12/1415/12/16

Fingerprint

Activity Recognition
Trajectories
Descriptors
Trajectory
Evaluation
Sports
Computer Vision
Surgery
Computer vision
Texture
Textures
Human
Framework
Vision
Traffic
Motion
Evaluate
Experimental Results
Experiment
Experiments

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Kataoka, H., Aoki, Y., Iwata, K., & Satoh, Y. (2015). Evaluation of vision-based human activity recognition in dense trajectory framework. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 9474, pp. 634-646). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9474). Springer Verlag. https://doi.org/10.1007/978-3-319-27857-5_57

Evaluation of vision-based human activity recognition in dense trajectory framework. / Kataoka, Hirokatsu; Aoki, Yoshimitsu; Iwata, Kenji; Satoh, Yutaka.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9474 Springer Verlag, 2015. p. 634-646 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9474).

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

Kataoka, H, Aoki, Y, Iwata, K & Satoh, Y 2015, Evaluation of vision-based human activity recognition in dense trajectory framework. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 9474, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9474, Springer Verlag, pp. 634-646, 11th International Symposium on Advances in Visual Computing, ISVC 2015, Las Vegas, United States, 15/12/14. https://doi.org/10.1007/978-3-319-27857-5_57
Kataoka H, Aoki Y, Iwata K, Satoh Y. Evaluation of vision-based human activity recognition in dense trajectory framework. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9474. Springer Verlag. 2015. p. 634-646. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-27857-5_57
Kataoka, Hirokatsu ; Aoki, Yoshimitsu ; Iwata, Kenji ; Satoh, Yutaka. / Evaluation of vision-based human activity recognition in dense trajectory framework. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 9474 Springer Verlag, 2015. pp. 634-646 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{3db3325f99d24e1bba6b0e4023e62791,
title = "Evaluation of vision-based human activity recognition in dense trajectory framework",
abstract = "Activity recognition has been an active research topic in computer vision. Recently, the most successful approaches use dense trajectories that extract a large number of trajectories and features on the trajectories into a codeword. In this paper, we evaluate various features in the framework of dense trajectories on several types of datasets. We implement 13 features in total by including five different types of descriptor, namely motion-, shape-, texture- trajectory- and co-occurrence-based feature descriptors. The experimental results show a relationship between feature descriptors and performance rate at each dataset. Different scenes of traffic, surgery, daily living and sports are used to analyze the feature characteristics. Moreover, we test how much the performance rate of concatenated vectors depends on the type, top-ranked in experiment and all 13 feature descriptors on fine-grained datasets. Feature evaluation is beneficial not only in the activity recognition problem, but also in other domains in spatio-temporal recognition.",
author = "Hirokatsu Kataoka and Yoshimitsu Aoki and Kenji Iwata and Yutaka Satoh",
year = "2015",
doi = "10.1007/978-3-319-27857-5_57",
language = "English",
isbn = "9783319278568",
volume = "9474",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "634--646",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Evaluation of vision-based human activity recognition in dense trajectory framework

AU - Kataoka, Hirokatsu

AU - Aoki, Yoshimitsu

AU - Iwata, Kenji

AU - Satoh, Yutaka

PY - 2015

Y1 - 2015

N2 - Activity recognition has been an active research topic in computer vision. Recently, the most successful approaches use dense trajectories that extract a large number of trajectories and features on the trajectories into a codeword. In this paper, we evaluate various features in the framework of dense trajectories on several types of datasets. We implement 13 features in total by including five different types of descriptor, namely motion-, shape-, texture- trajectory- and co-occurrence-based feature descriptors. The experimental results show a relationship between feature descriptors and performance rate at each dataset. Different scenes of traffic, surgery, daily living and sports are used to analyze the feature characteristics. Moreover, we test how much the performance rate of concatenated vectors depends on the type, top-ranked in experiment and all 13 feature descriptors on fine-grained datasets. Feature evaluation is beneficial not only in the activity recognition problem, but also in other domains in spatio-temporal recognition.

AB - Activity recognition has been an active research topic in computer vision. Recently, the most successful approaches use dense trajectories that extract a large number of trajectories and features on the trajectories into a codeword. In this paper, we evaluate various features in the framework of dense trajectories on several types of datasets. We implement 13 features in total by including five different types of descriptor, namely motion-, shape-, texture- trajectory- and co-occurrence-based feature descriptors. The experimental results show a relationship between feature descriptors and performance rate at each dataset. Different scenes of traffic, surgery, daily living and sports are used to analyze the feature characteristics. Moreover, we test how much the performance rate of concatenated vectors depends on the type, top-ranked in experiment and all 13 feature descriptors on fine-grained datasets. Feature evaluation is beneficial not only in the activity recognition problem, but also in other domains in spatio-temporal recognition.

UR - http://www.scopus.com/inward/record.url?scp=84952673860&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84952673860&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-27857-5_57

DO - 10.1007/978-3-319-27857-5_57

M3 - Conference contribution

SN - 9783319278568

VL - 9474

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

SP - 634

EP - 646

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

PB - Springer Verlag

ER -