Action attribute detection from sports videos with contextual constraints

Xiaodong Yu, Ching Lik Teo, Yezhou Yang, Cornelia Fermüller, Yiannis Aloimonos

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

1 Citation (Scopus)

Abstract

In this paper, we are interested in detecting action attributes from sports videos for event understanding and video analysis. Action attribute is a middle layer between low level motion features and high level action classes, which includes various motion patterns of human limbs and bodies and the interaction between human and objects. Successfully detecting action attributes provides a richer video description that facilitates many other important tasks, such action classification, video understanding, automatic video transcript, etc. A naive approach to deal with this challenging problem is to train a classifier for each attribute and then use them to detect attributes in novel videos independently. However, this independence assumption is often too strong, and as we show in our experiments, produces a large number of false positives in practice. We propose a novel approach that incorporates the contextual constraints for activity attribute detection. The temporal contexts within an attribute and the co-occurrence contexts between different attributes are modelled by a factorial conditional random field, which encourages agreement between different time points and attributes. The effectiveness of our methods are clearly illustrated by the experimental evaluations.

Original languageEnglish (US)
Title of host publicationBMVC 2013 - Electronic Proceedings of the British Machine Vision Conference 2013
PublisherBritish Machine Vision Association, BMVA
DOIs
StatePublished - 2013
Externally publishedYes
Event2013 24th British Machine Vision Conference, BMVC 2013 - Bristol, United Kingdom
Duration: Sep 9 2013Sep 13 2013

Other

Other2013 24th British Machine Vision Conference, BMVC 2013
CountryUnited Kingdom
CityBristol
Period9/9/139/13/13

Fingerprint

Sports
Classifiers
Experiments

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition

Cite this

Yu, X., Teo, C. L., Yang, Y., Fermüller, C., & Aloimonos, Y. (2013). Action attribute detection from sports videos with contextual constraints. In BMVC 2013 - Electronic Proceedings of the British Machine Vision Conference 2013 British Machine Vision Association, BMVA. https://doi.org/10.5244/C.27.79

Action attribute detection from sports videos with contextual constraints. / Yu, Xiaodong; Teo, Ching Lik; Yang, Yezhou; Fermüller, Cornelia; Aloimonos, Yiannis.

BMVC 2013 - Electronic Proceedings of the British Machine Vision Conference 2013. British Machine Vision Association, BMVA, 2013.

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

Yu, X, Teo, CL, Yang, Y, Fermüller, C & Aloimonos, Y 2013, Action attribute detection from sports videos with contextual constraints. in BMVC 2013 - Electronic Proceedings of the British Machine Vision Conference 2013. British Machine Vision Association, BMVA, 2013 24th British Machine Vision Conference, BMVC 2013, Bristol, United Kingdom, 9/9/13. https://doi.org/10.5244/C.27.79
Yu X, Teo CL, Yang Y, Fermüller C, Aloimonos Y. Action attribute detection from sports videos with contextual constraints. In BMVC 2013 - Electronic Proceedings of the British Machine Vision Conference 2013. British Machine Vision Association, BMVA. 2013 https://doi.org/10.5244/C.27.79
Yu, Xiaodong ; Teo, Ching Lik ; Yang, Yezhou ; Fermüller, Cornelia ; Aloimonos, Yiannis. / Action attribute detection from sports videos with contextual constraints. BMVC 2013 - Electronic Proceedings of the British Machine Vision Conference 2013. British Machine Vision Association, BMVA, 2013.
@inproceedings{5cc3252a03f740e68f9e727162c9f66c,
title = "Action attribute detection from sports videos with contextual constraints",
abstract = "In this paper, we are interested in detecting action attributes from sports videos for event understanding and video analysis. Action attribute is a middle layer between low level motion features and high level action classes, which includes various motion patterns of human limbs and bodies and the interaction between human and objects. Successfully detecting action attributes provides a richer video description that facilitates many other important tasks, such action classification, video understanding, automatic video transcript, etc. A naive approach to deal with this challenging problem is to train a classifier for each attribute and then use them to detect attributes in novel videos independently. However, this independence assumption is often too strong, and as we show in our experiments, produces a large number of false positives in practice. We propose a novel approach that incorporates the contextual constraints for activity attribute detection. The temporal contexts within an attribute and the co-occurrence contexts between different attributes are modelled by a factorial conditional random field, which encourages agreement between different time points and attributes. The effectiveness of our methods are clearly illustrated by the experimental evaluations.",
author = "Xiaodong Yu and Teo, {Ching Lik} and Yezhou Yang and Cornelia Ferm{\"u}ller and Yiannis Aloimonos",
year = "2013",
doi = "10.5244/C.27.79",
language = "English (US)",
booktitle = "BMVC 2013 - Electronic Proceedings of the British Machine Vision Conference 2013",
publisher = "British Machine Vision Association, BMVA",

}

TY - GEN

T1 - Action attribute detection from sports videos with contextual constraints

AU - Yu, Xiaodong

AU - Teo, Ching Lik

AU - Yang, Yezhou

AU - Fermüller, Cornelia

AU - Aloimonos, Yiannis

PY - 2013

Y1 - 2013

N2 - In this paper, we are interested in detecting action attributes from sports videos for event understanding and video analysis. Action attribute is a middle layer between low level motion features and high level action classes, which includes various motion patterns of human limbs and bodies and the interaction between human and objects. Successfully detecting action attributes provides a richer video description that facilitates many other important tasks, such action classification, video understanding, automatic video transcript, etc. A naive approach to deal with this challenging problem is to train a classifier for each attribute and then use them to detect attributes in novel videos independently. However, this independence assumption is often too strong, and as we show in our experiments, produces a large number of false positives in practice. We propose a novel approach that incorporates the contextual constraints for activity attribute detection. The temporal contexts within an attribute and the co-occurrence contexts between different attributes are modelled by a factorial conditional random field, which encourages agreement between different time points and attributes. The effectiveness of our methods are clearly illustrated by the experimental evaluations.

AB - In this paper, we are interested in detecting action attributes from sports videos for event understanding and video analysis. Action attribute is a middle layer between low level motion features and high level action classes, which includes various motion patterns of human limbs and bodies and the interaction between human and objects. Successfully detecting action attributes provides a richer video description that facilitates many other important tasks, such action classification, video understanding, automatic video transcript, etc. A naive approach to deal with this challenging problem is to train a classifier for each attribute and then use them to detect attributes in novel videos independently. However, this independence assumption is often too strong, and as we show in our experiments, produces a large number of false positives in practice. We propose a novel approach that incorporates the contextual constraints for activity attribute detection. The temporal contexts within an attribute and the co-occurrence contexts between different attributes are modelled by a factorial conditional random field, which encourages agreement between different time points and attributes. The effectiveness of our methods are clearly illustrated by the experimental evaluations.

UR - http://www.scopus.com/inward/record.url?scp=84898428895&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84898428895&partnerID=8YFLogxK

U2 - 10.5244/C.27.79

DO - 10.5244/C.27.79

M3 - Conference contribution

BT - BMVC 2013 - Electronic Proceedings of the British Machine Vision Conference 2013

PB - British Machine Vision Association, BMVA

ER -