Active planning, sensing, and recognition using a resource-constrained discriminant POMDP

Zhaowen Wang, Zhangyang Wang, Mark Moll, Po Sen Huang, Devin Grady, Nasser Nasrabadi, Thomas S Huang, Lydia Kavraki, Mark Allan Hasegawa-Johnson

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

Abstract

In this paper, we address the problem of object class recognition via observations from actively selected views/modalities/features under limited resource budgets. A Partially Observable Markov Decision Process (POMDP) is employed to find optimal sensing and recognition actions with the goal of long-term classification accuracy. Heterogeneous resource constraints-such as motion, number of measurements and bandwidth-are explicitly modeled in the state variable, and a prohibitively high penalty is used to prevent the violation of any resource constraint. To improve recognition performance, we further incorporate discriminative classification models with POMDP, and customize the reward function and observation model correspondingly. The proposed model is validated on several data sets for multi-view, multi-modal vehicle classification and multi-view face recognition, and demonstrates improvement in both recognition and resource management over greedy methods and previous POMDP formulations.

Original languageEnglish (US)
Title of host publicationProceedings - 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2014
PublisherIEEE Computer Society
Pages754-761
Number of pages8
ISBN (Electronic)9781479943098, 9781479943098
DOIs
StatePublished - Sep 24 2014
Event2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2014 - Columbus, United States
Duration: Jun 23 2014Jun 28 2014

Publication series

NameIEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops
ISSN (Print)2160-7508
ISSN (Electronic)2160-7516

Other

Other2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2014
CountryUnited States
CityColumbus
Period6/23/146/28/14

Fingerprint

Planning
Face recognition
Bandwidth

ASJC Scopus subject areas

  • Computer Vision and Pattern Recognition
  • Electrical and Electronic Engineering

Cite this

Wang, Z., Wang, Z., Moll, M., Huang, P. S., Grady, D., Nasrabadi, N., ... Hasegawa-Johnson, M. A. (2014). Active planning, sensing, and recognition using a resource-constrained discriminant POMDP. In Proceedings - 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2014 (pp. 754-761). [6910067] (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops). IEEE Computer Society. https://doi.org/10.1109/CVPRW.2014.116

Active planning, sensing, and recognition using a resource-constrained discriminant POMDP. / Wang, Zhaowen; Wang, Zhangyang; Moll, Mark; Huang, Po Sen; Grady, Devin; Nasrabadi, Nasser; Huang, Thomas S; Kavraki, Lydia; Hasegawa-Johnson, Mark Allan.

Proceedings - 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2014. IEEE Computer Society, 2014. p. 754-761 6910067 (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops).

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

Wang, Z, Wang, Z, Moll, M, Huang, PS, Grady, D, Nasrabadi, N, Huang, TS, Kavraki, L & Hasegawa-Johnson, MA 2014, Active planning, sensing, and recognition using a resource-constrained discriminant POMDP. in Proceedings - 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2014., 6910067, IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops, IEEE Computer Society, pp. 754-761, 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2014, Columbus, United States, 6/23/14. https://doi.org/10.1109/CVPRW.2014.116
Wang Z, Wang Z, Moll M, Huang PS, Grady D, Nasrabadi N et al. Active planning, sensing, and recognition using a resource-constrained discriminant POMDP. In Proceedings - 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2014. IEEE Computer Society. 2014. p. 754-761. 6910067. (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops). https://doi.org/10.1109/CVPRW.2014.116
Wang, Zhaowen ; Wang, Zhangyang ; Moll, Mark ; Huang, Po Sen ; Grady, Devin ; Nasrabadi, Nasser ; Huang, Thomas S ; Kavraki, Lydia ; Hasegawa-Johnson, Mark Allan. / Active planning, sensing, and recognition using a resource-constrained discriminant POMDP. Proceedings - 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2014. IEEE Computer Society, 2014. pp. 754-761 (IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops).
@inproceedings{f940d63cf93f43e39fa1db4ad0532b83,
title = "Active planning, sensing, and recognition using a resource-constrained discriminant POMDP",
abstract = "In this paper, we address the problem of object class recognition via observations from actively selected views/modalities/features under limited resource budgets. A Partially Observable Markov Decision Process (POMDP) is employed to find optimal sensing and recognition actions with the goal of long-term classification accuracy. Heterogeneous resource constraints-such as motion, number of measurements and bandwidth-are explicitly modeled in the state variable, and a prohibitively high penalty is used to prevent the violation of any resource constraint. To improve recognition performance, we further incorporate discriminative classification models with POMDP, and customize the reward function and observation model correspondingly. The proposed model is validated on several data sets for multi-view, multi-modal vehicle classification and multi-view face recognition, and demonstrates improvement in both recognition and resource management over greedy methods and previous POMDP formulations.",
author = "Zhaowen Wang and Zhangyang Wang and Mark Moll and Huang, {Po Sen} and Devin Grady and Nasser Nasrabadi and Huang, {Thomas S} and Lydia Kavraki and Hasegawa-Johnson, {Mark Allan}",
year = "2014",
month = "9",
day = "24",
doi = "10.1109/CVPRW.2014.116",
language = "English (US)",
series = "IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops",
publisher = "IEEE Computer Society",
pages = "754--761",
booktitle = "Proceedings - 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2014",

}

TY - GEN

T1 - Active planning, sensing, and recognition using a resource-constrained discriminant POMDP

AU - Wang, Zhaowen

AU - Wang, Zhangyang

AU - Moll, Mark

AU - Huang, Po Sen

AU - Grady, Devin

AU - Nasrabadi, Nasser

AU - Huang, Thomas S

AU - Kavraki, Lydia

AU - Hasegawa-Johnson, Mark Allan

PY - 2014/9/24

Y1 - 2014/9/24

N2 - In this paper, we address the problem of object class recognition via observations from actively selected views/modalities/features under limited resource budgets. A Partially Observable Markov Decision Process (POMDP) is employed to find optimal sensing and recognition actions with the goal of long-term classification accuracy. Heterogeneous resource constraints-such as motion, number of measurements and bandwidth-are explicitly modeled in the state variable, and a prohibitively high penalty is used to prevent the violation of any resource constraint. To improve recognition performance, we further incorporate discriminative classification models with POMDP, and customize the reward function and observation model correspondingly. The proposed model is validated on several data sets for multi-view, multi-modal vehicle classification and multi-view face recognition, and demonstrates improvement in both recognition and resource management over greedy methods and previous POMDP formulations.

AB - In this paper, we address the problem of object class recognition via observations from actively selected views/modalities/features under limited resource budgets. A Partially Observable Markov Decision Process (POMDP) is employed to find optimal sensing and recognition actions with the goal of long-term classification accuracy. Heterogeneous resource constraints-such as motion, number of measurements and bandwidth-are explicitly modeled in the state variable, and a prohibitively high penalty is used to prevent the violation of any resource constraint. To improve recognition performance, we further incorporate discriminative classification models with POMDP, and customize the reward function and observation model correspondingly. The proposed model is validated on several data sets for multi-view, multi-modal vehicle classification and multi-view face recognition, and demonstrates improvement in both recognition and resource management over greedy methods and previous POMDP formulations.

UR - http://www.scopus.com/inward/record.url?scp=84908506348&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84908506348&partnerID=8YFLogxK

U2 - 10.1109/CVPRW.2014.116

DO - 10.1109/CVPRW.2014.116

M3 - Conference contribution

AN - SCOPUS:84908506348

T3 - IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops

SP - 754

EP - 761

BT - Proceedings - 2014 IEEE Conference on Computer Vision and Pattern Recognition Workshops, CVPRW 2014

PB - IEEE Computer Society

ER -