Improving a Deep Learning based RGB-D Object Recognition Model by Ensemble Learning

Andreas Aakerberg, Kamal Nasrollahi, Thomas Heder

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

3 Citationer (Scopus)
354 Downloads (Pure)

Resumé

Augmenting RGB images with depth information is a well-known method to significantly improve the recognition accuracy of object recognition models. Another method to im- prove the performance of visual recognition models is ensemble learning. However, this method has not been widely explored in combination with deep convolutional neural network based RGB-D object recognition models. Hence, in this paper, we form different ensembles of complementary deep convolutional neural network models, and show that this can be used to increase the recognition performance beyond existing limits. Experiments on the Washington RGB-D Object Dataset show that our best performing ensemble improves the recognition performance with 0.7% compared to using the baseline model alone.
OriginalsprogEngelsk
Titel2017 Seventh International Conference on Image Processing Theory, Tools and Applications (IPTA)
Antal sider6
ForlagIEEE
Publikationsdato8 mar. 2018
Sider1-6
Artikelnummer8310101
ISBN (Trykt)978-1-5386-1843-1
ISBN (Elektronisk)978-1-5386-1842-4
DOI
StatusUdgivet - 8 mar. 2018
BegivenhedInternational Conference on Image Processing Theory, Tools and Applications - Montreal, Canada
Varighed: 28 nov. 20171 dec. 2017

Konference

KonferenceInternational Conference on Image Processing Theory, Tools and Applications
LandCanada
ByMontreal
Periode28/11/201701/12/2017
NavnInternational Conference on Image Processing Theory, Tools and Applications (IPTA)
ISSN2154-512X

Fingerprint

Object recognition
Neural networks
Deep learning
Experiments

Citer dette

Aakerberg, A., Nasrollahi, K., & Heder, T. (2018). Improving a Deep Learning based RGB-D Object Recognition Model by Ensemble Learning. I 2017 Seventh International Conference on Image Processing Theory, Tools and Applications (IPTA) (s. 1-6). [8310101] IEEE. International Conference on Image Processing Theory, Tools and Applications (IPTA) https://doi.org/10.1109/IPTA.2017.8310101
Aakerberg, Andreas ; Nasrollahi, Kamal ; Heder, Thomas. / Improving a Deep Learning based RGB-D Object Recognition Model by Ensemble Learning. 2017 Seventh International Conference on Image Processing Theory, Tools and Applications (IPTA). IEEE, 2018. s. 1-6 (International Conference on Image Processing Theory, Tools and Applications (IPTA)).
@inproceedings{59ad60be27db438a8f2e0ea40207070c,
title = "Improving a Deep Learning based RGB-D Object Recognition Model by Ensemble Learning",
abstract = "Augmenting RGB images with depth information is a well-known method to significantly improve the recognition accuracy of object recognition models. Another method to im- prove the performance of visual recognition models is ensemble learning. However, this method has not been widely explored in combination with deep convolutional neural network based RGB-D object recognition models. Hence, in this paper, we form different ensembles of complementary deep convolutional neural network models, and show that this can be used to increase the recognition performance beyond existing limits. Experiments on the Washington RGB-D Object Dataset show that our best performing ensemble improves the recognition performance with 0.7{\%} compared to using the baseline model alone.",
keywords = "Computer Vision, Convolutional Neural Networks, Deep Learning, Ensemble Learning, RGB-D",
author = "Andreas Aakerberg and Kamal Nasrollahi and Thomas Heder",
year = "2018",
month = "3",
day = "8",
doi = "10.1109/IPTA.2017.8310101",
language = "English",
isbn = "978-1-5386-1843-1",
series = "International Conference on Image Processing Theory, Tools and Applications (IPTA)",
pages = "1--6",
booktitle = "2017 Seventh International Conference on Image Processing Theory, Tools and Applications (IPTA)",
publisher = "IEEE",
address = "United States",

}

Aakerberg, A, Nasrollahi, K & Heder, T 2018, Improving a Deep Learning based RGB-D Object Recognition Model by Ensemble Learning. i 2017 Seventh International Conference on Image Processing Theory, Tools and Applications (IPTA)., 8310101, IEEE, International Conference on Image Processing Theory, Tools and Applications (IPTA), s. 1-6, International Conference on Image Processing Theory, Tools and Applications, Montreal, Canada, 28/11/2017. https://doi.org/10.1109/IPTA.2017.8310101

Improving a Deep Learning based RGB-D Object Recognition Model by Ensemble Learning. / Aakerberg, Andreas; Nasrollahi, Kamal; Heder, Thomas.

2017 Seventh International Conference on Image Processing Theory, Tools and Applications (IPTA). IEEE, 2018. s. 1-6 8310101 (International Conference on Image Processing Theory, Tools and Applications (IPTA)).

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

TY - GEN

T1 - Improving a Deep Learning based RGB-D Object Recognition Model by Ensemble Learning

AU - Aakerberg, Andreas

AU - Nasrollahi, Kamal

AU - Heder, Thomas

PY - 2018/3/8

Y1 - 2018/3/8

N2 - Augmenting RGB images with depth information is a well-known method to significantly improve the recognition accuracy of object recognition models. Another method to im- prove the performance of visual recognition models is ensemble learning. However, this method has not been widely explored in combination with deep convolutional neural network based RGB-D object recognition models. Hence, in this paper, we form different ensembles of complementary deep convolutional neural network models, and show that this can be used to increase the recognition performance beyond existing limits. Experiments on the Washington RGB-D Object Dataset show that our best performing ensemble improves the recognition performance with 0.7% compared to using the baseline model alone.

AB - Augmenting RGB images with depth information is a well-known method to significantly improve the recognition accuracy of object recognition models. Another method to im- prove the performance of visual recognition models is ensemble learning. However, this method has not been widely explored in combination with deep convolutional neural network based RGB-D object recognition models. Hence, in this paper, we form different ensembles of complementary deep convolutional neural network models, and show that this can be used to increase the recognition performance beyond existing limits. Experiments on the Washington RGB-D Object Dataset show that our best performing ensemble improves the recognition performance with 0.7% compared to using the baseline model alone.

KW - Computer Vision

KW - Convolutional Neural Networks

KW - Deep Learning

KW - Ensemble Learning

KW - RGB-D

UR - http://www.scopus.com/inward/record.url?scp=85050615657&partnerID=8YFLogxK

U2 - 10.1109/IPTA.2017.8310101

DO - 10.1109/IPTA.2017.8310101

M3 - Article in proceeding

SN - 978-1-5386-1843-1

T3 - International Conference on Image Processing Theory, Tools and Applications (IPTA)

SP - 1

EP - 6

BT - 2017 Seventh International Conference on Image Processing Theory, Tools and Applications (IPTA)

PB - IEEE

ER -

Aakerberg A, Nasrollahi K, Heder T. Improving a Deep Learning based RGB-D Object Recognition Model by Ensemble Learning. I 2017 Seventh International Conference on Image Processing Theory, Tools and Applications (IPTA). IEEE. 2018. s. 1-6. 8310101. (International Conference on Image Processing Theory, Tools and Applications (IPTA)). https://doi.org/10.1109/IPTA.2017.8310101