Keep it Accurate and Diverse: Enhancing Action Recognition Performance by Ensemble Learning

Mohammad Ali Bagheri , Qigang Gao , Sergio Escalera Guerrero, Albert Clapés, Kamal Nasrollahi, Michael Boelstoft Holte, Thomas B. Moeslund

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceartikel i proceedingForskningpeer review

7 Citationer (Scopus)
486 Downloads (Pure)

Resumé

The performance of different action recognition tech- niques has recently been studied by several computer vision researchers. However, the potential improvement in classi- fication through classifier fusion by ensemble-based meth- ods has remained unattended. In this work, we evaluate the performance of an ensemble of action learning techniques, each performing the recognition task from a different per- spective. The underlying idea is that instead of aiming a very sophisticated and powerful representation/learning technique, we can learn action categories using a set of relatively simple and diverse classifiers, each trained with different feature set. In addition, combining the outputs of several learners can reduce the risk of an unfortunate selec- tion of a learner on an unseen action recognition scenario. This leads to having a more robust and general-applicable framework. In order to improve the recognition perfor- mance, a powerful combination strategy is utilized based on the Dempster-Shafer theory, which can effectively make use of diversity of base learners trained on different sources of information. The recognition results of the individual clas- sifiers are compared with those obtained from fusing the classifiers’ output, showing enhanced performance of the proposed methodology.
OriginalsprogEngelsk
TitelIEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPRW), 2015
ForlagIEEE Computer Society Press
Publikationsdato7 jun. 2015
Sider22-29
ISBN (Trykt)978-1-4673-6759-2
DOI
StatusUdgivet - 7 jun. 2015
Begivenhed2015 IEEE Conference on Computer Vision and Pattern Recognition Workshop: Looking at People - Boston, USA
Varighed: 6 jun. 201512 jun. 2015

Konference

Konference2015 IEEE Conference on Computer Vision and Pattern Recognition Workshop
LandUSA
ByBoston
Periode06/06/201512/06/2015

Fingerprint

Classifiers
Computer vision
Fusion reactions

Citer dette

Ali Bagheri , M., Gao , Q., Guerrero, S. E., Clapés, A., Nasrollahi, K., Holte, M. B., & Moeslund, T. B. (2015). Keep it Accurate and Diverse: Enhancing Action Recognition Performance by Ensemble Learning. I IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPRW), 2015 (s. 22-29). IEEE Computer Society Press. https://doi.org/10.1109/CVPRW.2015.7301332
Ali Bagheri , Mohammad ; Gao , Qigang ; Guerrero, Sergio Escalera ; Clapés, Albert ; Nasrollahi, Kamal ; Holte, Michael Boelstoft ; Moeslund, Thomas B. / Keep it Accurate and Diverse : Enhancing Action Recognition Performance by Ensemble Learning. IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPRW), 2015. IEEE Computer Society Press, 2015. s. 22-29
@inproceedings{cbdb26a5c3684bed87fe2cc0dbec2170,
title = "Keep it Accurate and Diverse: Enhancing Action Recognition Performance by Ensemble Learning",
abstract = "The performance of different action recognition tech- niques has recently been studied by several computer vision researchers. However, the potential improvement in classi- fication through classifier fusion by ensemble-based meth- ods has remained unattended. In this work, we evaluate the performance of an ensemble of action learning techniques, each performing the recognition task from a different per- spective. The underlying idea is that instead of aiming a very sophisticated and powerful representation/learning technique, we can learn action categories using a set of relatively simple and diverse classifiers, each trained with different feature set. In addition, combining the outputs of several learners can reduce the risk of an unfortunate selec- tion of a learner on an unseen action recognition scenario. This leads to having a more robust and general-applicable framework. In order to improve the recognition perfor- mance, a powerful combination strategy is utilized based on the Dempster-Shafer theory, which can effectively make use of diversity of base learners trained on different sources of information. The recognition results of the individual clas- sifiers are compared with those obtained from fusing the classifiers’ output, showing enhanced performance of the proposed methodology.",
author = "{Ali Bagheri}, Mohammad and Qigang Gao and Guerrero, {Sergio Escalera} and Albert Clap{\'e}s and Kamal Nasrollahi and Holte, {Michael Boelstoft} and Moeslund, {Thomas B.}",
year = "2015",
month = "6",
day = "7",
doi = "10.1109/CVPRW.2015.7301332",
language = "English",
isbn = "978-1-4673-6759-2",
pages = "22--29",
booktitle = "IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPRW), 2015",
publisher = "IEEE Computer Society Press",
address = "United States",

}

Ali Bagheri , M, Gao , Q, Guerrero, SE, Clapés, A, Nasrollahi, K, Holte, MB & Moeslund, TB 2015, Keep it Accurate and Diverse: Enhancing Action Recognition Performance by Ensemble Learning. i IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPRW), 2015. IEEE Computer Society Press, s. 22-29, 2015 IEEE Conference on Computer Vision and Pattern Recognition Workshop, Boston, USA, 06/06/2015. https://doi.org/10.1109/CVPRW.2015.7301332

Keep it Accurate and Diverse : Enhancing Action Recognition Performance by Ensemble Learning. / Ali Bagheri , Mohammad; Gao , Qigang; Guerrero, Sergio Escalera; Clapés, Albert; Nasrollahi, Kamal; Holte, Michael Boelstoft; Moeslund, Thomas B.

IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPRW), 2015. IEEE Computer Society Press, 2015. s. 22-29.

Publikation: Bidrag til bog/antologi/rapport/konference proceedingKonferenceartikel i proceedingForskningpeer review

TY - GEN

T1 - Keep it Accurate and Diverse

T2 - Enhancing Action Recognition Performance by Ensemble Learning

AU - Ali Bagheri , Mohammad

AU - Gao , Qigang

AU - Guerrero, Sergio Escalera

AU - Clapés, Albert

AU - Nasrollahi, Kamal

AU - Holte, Michael Boelstoft

AU - Moeslund, Thomas B.

PY - 2015/6/7

Y1 - 2015/6/7

N2 - The performance of different action recognition tech- niques has recently been studied by several computer vision researchers. However, the potential improvement in classi- fication through classifier fusion by ensemble-based meth- ods has remained unattended. In this work, we evaluate the performance of an ensemble of action learning techniques, each performing the recognition task from a different per- spective. The underlying idea is that instead of aiming a very sophisticated and powerful representation/learning technique, we can learn action categories using a set of relatively simple and diverse classifiers, each trained with different feature set. In addition, combining the outputs of several learners can reduce the risk of an unfortunate selec- tion of a learner on an unseen action recognition scenario. This leads to having a more robust and general-applicable framework. In order to improve the recognition perfor- mance, a powerful combination strategy is utilized based on the Dempster-Shafer theory, which can effectively make use of diversity of base learners trained on different sources of information. The recognition results of the individual clas- sifiers are compared with those obtained from fusing the classifiers’ output, showing enhanced performance of the proposed methodology.

AB - The performance of different action recognition tech- niques has recently been studied by several computer vision researchers. However, the potential improvement in classi- fication through classifier fusion by ensemble-based meth- ods has remained unattended. In this work, we evaluate the performance of an ensemble of action learning techniques, each performing the recognition task from a different per- spective. The underlying idea is that instead of aiming a very sophisticated and powerful representation/learning technique, we can learn action categories using a set of relatively simple and diverse classifiers, each trained with different feature set. In addition, combining the outputs of several learners can reduce the risk of an unfortunate selec- tion of a learner on an unseen action recognition scenario. This leads to having a more robust and general-applicable framework. In order to improve the recognition perfor- mance, a powerful combination strategy is utilized based on the Dempster-Shafer theory, which can effectively make use of diversity of base learners trained on different sources of information. The recognition results of the individual clas- sifiers are compared with those obtained from fusing the classifiers’ output, showing enhanced performance of the proposed methodology.

U2 - 10.1109/CVPRW.2015.7301332

DO - 10.1109/CVPRW.2015.7301332

M3 - Article in proceeding

SN - 978-1-4673-6759-2

SP - 22

EP - 29

BT - IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPRW), 2015

PB - IEEE Computer Society Press

ER -

Ali Bagheri M, Gao Q, Guerrero SE, Clapés A, Nasrollahi K, Holte MB et al. Keep it Accurate and Diverse: Enhancing Action Recognition Performance by Ensemble Learning. I IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPRW), 2015. IEEE Computer Society Press. 2015. s. 22-29 https://doi.org/10.1109/CVPRW.2015.7301332