Generic online animal activity recognition on collar tags

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

3 Citations (Scopus)
52 Downloads (Pure)

Abstract

Animal behaviour is a commonly-used and sensitive indicator of animal welfare. Moreover, the behaviour of animals can provide rich information about their environment. For online activity recognition on collar tags of animals, fundamental challenges include: limited energy resources, limited CPU and memory availability, and heterogeneity of animals. In this paper, we propose to tackle these challenges with a framework that employs Multitask Learning for embedded platforms. We train the classifiers with shared training data and a shared feature-representation. We show that Multitask Learning has a significant positive effect on the performance of the classifiers. Furthermore, we compare 7 types of classifiers in terms of resource usage and activity recognition performance on real-world movement data from goats and sheep. A Deep Neural Network could obtain an accuracy of 94% when tested with the data from both species. Our results show that a Deep Neural Network performs the best among the compared classifiers in terms of complexity versus performance. This work supports the development of a robust generic classifier that can run on a small embedded system with good performance, as well as sustain the lifetime of online activity recognition systems.
Original languageEnglish
Title of host publicationUbiComp'17
Subtitle of host publicationProceedings of the 2017 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2017 ACM International Symposium on Wearable Computers
Place of PublicationNew York, NY
PublisherACM Press
Pages597-606
Number of pages10
ISBN (Print)9781450351904
DOIs
Publication statusPublished - 2017
Event2017 ACM International Joint Conference on Pervasive and Ubiquitous Computing, UbiComp 2017 - Maui, United States
Duration: 11 Sep 201715 Sep 2017
http://ubicomp.org/ubicomp2017/

Conference

Conference2017 ACM International Joint Conference on Pervasive and Ubiquitous Computing, UbiComp 2017
Abbreviated titleUbiComp
CountryUnited States
CityMaui
Period11/09/1715/09/17
Internet address

Fingerprint

Animals
Classifiers
Energy resources
Embedded systems
Program processors
Availability
Data storage equipment
Deep neural networks

Cite this

Kamminga, J. W., Bisby, H. C., Le, D. V., Meratnia, N., & Havinga, P. J. M. (2017). Generic online animal activity recognition on collar tags. In UbiComp'17: Proceedings of the 2017 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2017 ACM International Symposium on Wearable Computers (pp. 597-606). New York, NY: ACM Press. https://doi.org/10.1145/3123024.3124407
Kamminga, Jacob W. ; Bisby, Helena C. ; Le, Duc V. ; Meratnia, Nirvana ; Havinga, Paul J.M. / Generic online animal activity recognition on collar tags. UbiComp'17: Proceedings of the 2017 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2017 ACM International Symposium on Wearable Computers. New York, NY : ACM Press, 2017. pp. 597-606
@inproceedings{442fb5c6c0124576a01aecce641e4ab4,
title = "Generic online animal activity recognition on collar tags",
abstract = "Animal behaviour is a commonly-used and sensitive indicator of animal welfare. Moreover, the behaviour of animals can provide rich information about their environment. For online activity recognition on collar tags of animals, fundamental challenges include: limited energy resources, limited CPU and memory availability, and heterogeneity of animals. In this paper, we propose to tackle these challenges with a framework that employs Multitask Learning for embedded platforms. We train the classifiers with shared training data and a shared feature-representation. We show that Multitask Learning has a significant positive effect on the performance of the classifiers. Furthermore, we compare 7 types of classifiers in terms of resource usage and activity recognition performance on real-world movement data from goats and sheep. A Deep Neural Network could obtain an accuracy of 94{\%} when tested with the data from both species. Our results show that a Deep Neural Network performs the best among the compared classifiers in terms of complexity versus performance. This work supports the development of a robust generic classifier that can run on a small embedded system with good performance, as well as sustain the lifetime of online activity recognition systems.",
author = "Kamminga, {Jacob W.} and Bisby, {Helena C.} and Le, {Duc V.} and Nirvana Meratnia and Havinga, {Paul J.M.}",
year = "2017",
doi = "10.1145/3123024.3124407",
language = "English",
isbn = "9781450351904",
pages = "597--606",
booktitle = "UbiComp'17",
publisher = "ACM Press",

}

Kamminga, JW, Bisby, HC, Le, DV, Meratnia, N & Havinga, PJM 2017, Generic online animal activity recognition on collar tags. in UbiComp'17: Proceedings of the 2017 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2017 ACM International Symposium on Wearable Computers. ACM Press, New York, NY, pp. 597-606, 2017 ACM International Joint Conference on Pervasive and Ubiquitous Computing, UbiComp 2017, Maui, United States, 11/09/17. https://doi.org/10.1145/3123024.3124407

Generic online animal activity recognition on collar tags. / Kamminga, Jacob W.; Bisby, Helena C.; Le, Duc V.; Meratnia, Nirvana ; Havinga, Paul J.M.

UbiComp'17: Proceedings of the 2017 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2017 ACM International Symposium on Wearable Computers. New York, NY : ACM Press, 2017. p. 597-606.

Research output: Chapter in Book/Report/Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - Generic online animal activity recognition on collar tags

AU - Kamminga, Jacob W.

AU - Bisby, Helena C.

AU - Le, Duc V.

AU - Meratnia, Nirvana

AU - Havinga, Paul J.M.

PY - 2017

Y1 - 2017

N2 - Animal behaviour is a commonly-used and sensitive indicator of animal welfare. Moreover, the behaviour of animals can provide rich information about their environment. For online activity recognition on collar tags of animals, fundamental challenges include: limited energy resources, limited CPU and memory availability, and heterogeneity of animals. In this paper, we propose to tackle these challenges with a framework that employs Multitask Learning for embedded platforms. We train the classifiers with shared training data and a shared feature-representation. We show that Multitask Learning has a significant positive effect on the performance of the classifiers. Furthermore, we compare 7 types of classifiers in terms of resource usage and activity recognition performance on real-world movement data from goats and sheep. A Deep Neural Network could obtain an accuracy of 94% when tested with the data from both species. Our results show that a Deep Neural Network performs the best among the compared classifiers in terms of complexity versus performance. This work supports the development of a robust generic classifier that can run on a small embedded system with good performance, as well as sustain the lifetime of online activity recognition systems.

AB - Animal behaviour is a commonly-used and sensitive indicator of animal welfare. Moreover, the behaviour of animals can provide rich information about their environment. For online activity recognition on collar tags of animals, fundamental challenges include: limited energy resources, limited CPU and memory availability, and heterogeneity of animals. In this paper, we propose to tackle these challenges with a framework that employs Multitask Learning for embedded platforms. We train the classifiers with shared training data and a shared feature-representation. We show that Multitask Learning has a significant positive effect on the performance of the classifiers. Furthermore, we compare 7 types of classifiers in terms of resource usage and activity recognition performance on real-world movement data from goats and sheep. A Deep Neural Network could obtain an accuracy of 94% when tested with the data from both species. Our results show that a Deep Neural Network performs the best among the compared classifiers in terms of complexity versus performance. This work supports the development of a robust generic classifier that can run on a small embedded system with good performance, as well as sustain the lifetime of online activity recognition systems.

U2 - 10.1145/3123024.3124407

DO - 10.1145/3123024.3124407

M3 - Conference contribution

SN - 9781450351904

SP - 597

EP - 606

BT - UbiComp'17

PB - ACM Press

CY - New York, NY

ER -

Kamminga JW, Bisby HC, Le DV, Meratnia N, Havinga PJM. Generic online animal activity recognition on collar tags. In UbiComp'17: Proceedings of the 2017 ACM International Joint Conference on Pervasive and Ubiquitous Computing and Proceedings of the 2017 ACM International Symposium on Wearable Computers. New York, NY: ACM Press. 2017. p. 597-606 https://doi.org/10.1145/3123024.3124407