Weakly supervised learning with convolutional neural networks for power line localization

Sang Jun Lee; Jong Pil Yun; Hyeyeon Choi; Wookyong Kwon; Gyogwon Koo; Sang Woo Kim

doi:10.1109/SSCI.2017.8285410

Weakly supervised learning with convolutional neural networks for power line localization

Sang Jun Lee, Jong Pil Yun, Hyeyeon Choi, Wookyong Kwon, Gyogwon Koo, Sang Woo Kim

Division of Intelligent Robot

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

28 Scopus citations

Abstract

Localization of power lines is important to monitor electricity infrastructures by using unmanned aerial vehicles. Although deep learning is a powerful method to solve computer vision problems, constructing pixel-level ground-truth data for object localization is an exhausting task. This paper proposes a weakly supervised learning algorithm for the localization of power lines by only using image-level class labels. The proposed algorithm classifies sub-regions by using a sliding window and convolutional neural network (CNN). A sub-region is filtered out if it is classified into an image without any power line. If a sub-region is classified into an image with a power line, then its feature maps of intermediate convolutional layers are combined to visualize the location of the power line. Experiments were conducted on actual aerial images to demonstrate the effectiveness of the proposed algorithm.

Original language	English
Title of host publication	2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	1-8
Number of pages	8
ISBN (Electronic)	9781538627259
DOIs	https://doi.org/10.1109/SSCI.2017.8285410
State	Published - 1 Jul 2017
Event	2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Honolulu, United States Duration: 27 Nov 2017 → 1 Dec 2017

Publication series

Name	2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings
Volume	2018-January

Conference

Conference	2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017
Country/Territory	United States
City	Honolulu
Period	27/11/17 → 1/12/17

Bibliographical note

Publisher Copyright:
© 2017 IEEE.

Keywords

Convolutional neural network
electricity transmission line
object localization
weakly supervised learning

Access to Document

10.1109/SSCI.2017.8285410

Cite this

Lee, S. J., Yun, J. P., Choi, H., Kwon, W., Koo, G., & Kim, S. W. (2017). Weakly supervised learning with convolutional neural networks for power line localization. In 2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings (pp. 1-8). (2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings; Vol. 2018-January). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/SSCI.2017.8285410

Lee, Sang Jun ; Yun, Jong Pil ; Choi, Hyeyeon et al. / Weakly supervised learning with convolutional neural networks for power line localization. 2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2017. pp. 1-8 (2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings).

@inproceedings{0af957eb0d444e1098393670f050b313,

title = "Weakly supervised learning with convolutional neural networks for power line localization",

abstract = "Localization of power lines is important to monitor electricity infrastructures by using unmanned aerial vehicles. Although deep learning is a powerful method to solve computer vision problems, constructing pixel-level ground-truth data for object localization is an exhausting task. This paper proposes a weakly supervised learning algorithm for the localization of power lines by only using image-level class labels. The proposed algorithm classifies sub-regions by using a sliding window and convolutional neural network (CNN). A sub-region is filtered out if it is classified into an image without any power line. If a sub-region is classified into an image with a power line, then its feature maps of intermediate convolutional layers are combined to visualize the location of the power line. Experiments were conducted on actual aerial images to demonstrate the effectiveness of the proposed algorithm.",

keywords = "Convolutional neural network, electricity transmission line, object localization, weakly supervised learning",

author = "Lee, {Sang Jun} and Yun, {Jong Pil} and Hyeyeon Choi and Wookyong Kwon and Gyogwon Koo and Kim, {Sang Woo}",

note = "Publisher Copyright: {\textcopyright} 2017 IEEE.; 2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 ; Conference date: 27-11-2017 Through 01-12-2017",

year = "2017",

month = jul,

day = "1",

doi = "10.1109/SSCI.2017.8285410",

language = "English",

series = "2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "1--8",

booktitle = "2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings",

}

Lee, SJ, Yun, JP, Choi, H, Kwon, W, Koo, G & Kim, SW 2017, Weakly supervised learning with convolutional neural networks for power line localization. in 2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings. 2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings, vol. 2018-January, Institute of Electrical and Electronics Engineers Inc., pp. 1-8, 2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017, Honolulu, United States, 27/11/17. https://doi.org/10.1109/SSCI.2017.8285410

Weakly supervised learning with convolutional neural networks for power line localization. / Lee, Sang Jun; Yun, Jong Pil; Choi, Hyeyeon et al.
2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2017. p. 1-8 (2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings; Vol. 2018-January).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Weakly supervised learning with convolutional neural networks for power line localization

AU - Lee, Sang Jun

AU - Yun, Jong Pil

AU - Choi, Hyeyeon

AU - Kwon, Wookyong

AU - Koo, Gyogwon

AU - Kim, Sang Woo

PY - 2017/7/1

Y1 - 2017/7/1

N2 - Localization of power lines is important to monitor electricity infrastructures by using unmanned aerial vehicles. Although deep learning is a powerful method to solve computer vision problems, constructing pixel-level ground-truth data for object localization is an exhausting task. This paper proposes a weakly supervised learning algorithm for the localization of power lines by only using image-level class labels. The proposed algorithm classifies sub-regions by using a sliding window and convolutional neural network (CNN). A sub-region is filtered out if it is classified into an image without any power line. If a sub-region is classified into an image with a power line, then its feature maps of intermediate convolutional layers are combined to visualize the location of the power line. Experiments were conducted on actual aerial images to demonstrate the effectiveness of the proposed algorithm.

AB - Localization of power lines is important to monitor electricity infrastructures by using unmanned aerial vehicles. Although deep learning is a powerful method to solve computer vision problems, constructing pixel-level ground-truth data for object localization is an exhausting task. This paper proposes a weakly supervised learning algorithm for the localization of power lines by only using image-level class labels. The proposed algorithm classifies sub-regions by using a sliding window and convolutional neural network (CNN). A sub-region is filtered out if it is classified into an image without any power line. If a sub-region is classified into an image with a power line, then its feature maps of intermediate convolutional layers are combined to visualize the location of the power line. Experiments were conducted on actual aerial images to demonstrate the effectiveness of the proposed algorithm.

KW - Convolutional neural network

KW - electricity transmission line

KW - object localization

KW - weakly supervised learning

UR - http://www.scopus.com/inward/record.url?scp=85046136401&partnerID=8YFLogxK

U2 - 10.1109/SSCI.2017.8285410

DO - 10.1109/SSCI.2017.8285410

M3 - Conference contribution

AN - SCOPUS:85046136401

T3 - 2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings

SP - 1

EP - 8

BT - 2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017

Y2 - 27 November 2017 through 1 December 2017

ER -

Lee SJ, Yun JP, Choi H, Kwon W, Koo G, Kim SW. Weakly supervised learning with convolutional neural networks for power line localization. In 2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2017. p. 1-8. (2017 IEEE Symposium Series on Computational Intelligence, SSCI 2017 - Proceedings). doi: 10.1109/SSCI.2017.8285410

Weakly supervised learning with convolutional neural networks for power line localization

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this