Building detection and segmentation using a CNN with automatically generated training data

Xiangyu Zhuo, Friedrich Fraundorfer, Franz Kurz, Peter Reinartz

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

Abstract

Significantly outperforming traditional machine learning methods, deep convolutional neural networks have gained increasing popularity in the application of image classification and segmentation. Nevertheless, deep learning-based methods usually require a large amount of training data, which is quite labor-intensive and time-demanding. To deal with the problem in generating training data, we propose in this paper a novel approach to generate image annotations by transferring labels from aerial images to UAV images and refine the annotations using a densely connected CRF model with an embedded naive Bayes classifier. The generated annotations not only present correct semantic labels, but also preserve accurate class boundaries. To validate the utility of these automatic annotations, we deploy them as training data for pixel-wise image segmentation and compare the results with the segmentation using manual annotations. Experiment results demonstrate that the automatic annotations can achieve comparable segmentation accuracy as the manual annotations.

Original languageEnglish
Title of host publication2018 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2018 - Proceedings
PublisherInstitute of Electrical and Electronics Engineers
Pages3461-3464
Number of pages4
ISBN (Electronic)9781538671504
DOIs
Publication statusPublished - 31 Oct 2018
Event38th Annual IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2018 - Valencia, Spain
Duration: 22 Jul 201827 Jul 2018

Publication series

NameInternational Geoscience and Remote Sensing Symposium (IGARSS)
Volume2018-July

Conference

Conference38th Annual IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2018
CountrySpain
CityValencia
Period22/07/1827/07/18

Fingerprint

Image segmentation
segmentation
Labels
Image classification
Unmanned aerial vehicles (UAV)
Learning systems
Classifiers
Pixels
Semantics
Personnel
Antennas
Neural networks
image classification
pixel
labor
learning
Experiments
detection
experiment
Deep learning

Keywords

  • Automatic image annotation
  • Deep learning
  • Image segmentation
  • Label propagation

ASJC Scopus subject areas

  • Computer Science Applications
  • Earth and Planetary Sciences(all)

Cite this

Zhuo, X., Fraundorfer, F., Kurz, F., & Reinartz, P. (2018). Building detection and segmentation using a CNN with automatically generated training data. In 2018 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2018 - Proceedings (pp. 3461-3464). [8518521] (International Geoscience and Remote Sensing Symposium (IGARSS); Vol. 2018-July). Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/IGARSS.2018.8518521

Building detection and segmentation using a CNN with automatically generated training data. / Zhuo, Xiangyu; Fraundorfer, Friedrich; Kurz, Franz; Reinartz, Peter.

2018 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2018 - Proceedings. Institute of Electrical and Electronics Engineers, 2018. p. 3461-3464 8518521 (International Geoscience and Remote Sensing Symposium (IGARSS); Vol. 2018-July).

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

Zhuo, X, Fraundorfer, F, Kurz, F & Reinartz, P 2018, Building detection and segmentation using a CNN with automatically generated training data. in 2018 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2018 - Proceedings., 8518521, International Geoscience and Remote Sensing Symposium (IGARSS), vol. 2018-July, Institute of Electrical and Electronics Engineers, pp. 3461-3464, 38th Annual IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2018, Valencia, Spain, 22/07/18. https://doi.org/10.1109/IGARSS.2018.8518521
Zhuo X, Fraundorfer F, Kurz F, Reinartz P. Building detection and segmentation using a CNN with automatically generated training data. In 2018 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2018 - Proceedings. Institute of Electrical and Electronics Engineers. 2018. p. 3461-3464. 8518521. (International Geoscience and Remote Sensing Symposium (IGARSS)). https://doi.org/10.1109/IGARSS.2018.8518521
Zhuo, Xiangyu ; Fraundorfer, Friedrich ; Kurz, Franz ; Reinartz, Peter. / Building detection and segmentation using a CNN with automatically generated training data. 2018 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2018 - Proceedings. Institute of Electrical and Electronics Engineers, 2018. pp. 3461-3464 (International Geoscience and Remote Sensing Symposium (IGARSS)).
@inproceedings{6090ca76f3d745af9175345ecd64df6f,
title = "Building detection and segmentation using a CNN with automatically generated training data",
abstract = "Significantly outperforming traditional machine learning methods, deep convolutional neural networks have gained increasing popularity in the application of image classification and segmentation. Nevertheless, deep learning-based methods usually require a large amount of training data, which is quite labor-intensive and time-demanding. To deal with the problem in generating training data, we propose in this paper a novel approach to generate image annotations by transferring labels from aerial images to UAV images and refine the annotations using a densely connected CRF model with an embedded naive Bayes classifier. The generated annotations not only present correct semantic labels, but also preserve accurate class boundaries. To validate the utility of these automatic annotations, we deploy them as training data for pixel-wise image segmentation and compare the results with the segmentation using manual annotations. Experiment results demonstrate that the automatic annotations can achieve comparable segmentation accuracy as the manual annotations.",
keywords = "Automatic image annotation, Deep learning, Image segmentation, Label propagation",
author = "Xiangyu Zhuo and Friedrich Fraundorfer and Franz Kurz and Peter Reinartz",
year = "2018",
month = "10",
day = "31",
doi = "10.1109/IGARSS.2018.8518521",
language = "English",
series = "International Geoscience and Remote Sensing Symposium (IGARSS)",
publisher = "Institute of Electrical and Electronics Engineers",
pages = "3461--3464",
booktitle = "2018 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2018 - Proceedings",
address = "United States",

}

TY - GEN

T1 - Building detection and segmentation using a CNN with automatically generated training data

AU - Zhuo, Xiangyu

AU - Fraundorfer, Friedrich

AU - Kurz, Franz

AU - Reinartz, Peter

PY - 2018/10/31

Y1 - 2018/10/31

N2 - Significantly outperforming traditional machine learning methods, deep convolutional neural networks have gained increasing popularity in the application of image classification and segmentation. Nevertheless, deep learning-based methods usually require a large amount of training data, which is quite labor-intensive and time-demanding. To deal with the problem in generating training data, we propose in this paper a novel approach to generate image annotations by transferring labels from aerial images to UAV images and refine the annotations using a densely connected CRF model with an embedded naive Bayes classifier. The generated annotations not only present correct semantic labels, but also preserve accurate class boundaries. To validate the utility of these automatic annotations, we deploy them as training data for pixel-wise image segmentation and compare the results with the segmentation using manual annotations. Experiment results demonstrate that the automatic annotations can achieve comparable segmentation accuracy as the manual annotations.

AB - Significantly outperforming traditional machine learning methods, deep convolutional neural networks have gained increasing popularity in the application of image classification and segmentation. Nevertheless, deep learning-based methods usually require a large amount of training data, which is quite labor-intensive and time-demanding. To deal with the problem in generating training data, we propose in this paper a novel approach to generate image annotations by transferring labels from aerial images to UAV images and refine the annotations using a densely connected CRF model with an embedded naive Bayes classifier. The generated annotations not only present correct semantic labels, but also preserve accurate class boundaries. To validate the utility of these automatic annotations, we deploy them as training data for pixel-wise image segmentation and compare the results with the segmentation using manual annotations. Experiment results demonstrate that the automatic annotations can achieve comparable segmentation accuracy as the manual annotations.

KW - Automatic image annotation

KW - Deep learning

KW - Image segmentation

KW - Label propagation

UR - http://www.scopus.com/inward/record.url?scp=85064182092&partnerID=8YFLogxK

U2 - 10.1109/IGARSS.2018.8518521

DO - 10.1109/IGARSS.2018.8518521

M3 - Conference contribution

T3 - International Geoscience and Remote Sensing Symposium (IGARSS)

SP - 3461

EP - 3464

BT - 2018 IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2018 - Proceedings

PB - Institute of Electrical and Electronics Engineers

ER -