Using convolutional neural nwtwork for signboard detection on street view images

Pin Xu Chen, Jiann-Yeou Rau

Research output: Contribution to conferencePaper

Abstract

In order to efficiently build and update store information in digital maps, a convolutional neural network (CNN) model called Faster R-CNN proposed in 2015 is used for signboard detection on street view images. Google's Inception-ResNet-v2 model is the feature extractor in our model and a series of fully-connected layers are used for classification and bounding box regression. In the beginning, a portion of street view images is labelled for training model. Then, the bounding boxes and corresponding probabilities of signboard detection results can be obtained by applying our model to the other portion of street view images. In the evaluation, the precision of our method based on CNN is about 94.87%. In additional evaluations, all the precisions are above 93% after respectively adding Gaussian noise, Gaussian blur, horizontal flip, and change of brightness to the testing images, which shows high potential of our model for future applications. For example, the change analysis or character recognition techniques can be applied to street view images acquired by a mobile mapping system for updating store's attribute as well as geographic location automatically.

Original languageEnglish
Pages1997-2004
Number of pages8
Publication statusPublished - 2018 Jan 1
Event39th Asian Conference on Remote Sensing: Remote Sensing Enabling Prosperity, ACRS 2018 - Kuala Lumpur, Malaysia
Duration: 2018 Oct 152018 Oct 19

Conference

Conference39th Asian Conference on Remote Sensing: Remote Sensing Enabling Prosperity, ACRS 2018
CountryMalaysia
CityKuala Lumpur
Period18-10-1518-10-19

Fingerprint

Neural networks
Character recognition
digital map
detection
Luminance
Testing
evaluation
analysis
attribute
method

All Science Journal Classification (ASJC) codes

  • Computer Science Applications
  • Information Systems
  • Earth and Planetary Sciences(all)
  • Computer Networks and Communications

Cite this

Chen, P. X., & Rau, J-Y. (2018). Using convolutional neural nwtwork for signboard detection on street view images. 1997-2004. Paper presented at 39th Asian Conference on Remote Sensing: Remote Sensing Enabling Prosperity, ACRS 2018, Kuala Lumpur, Malaysia.
Chen, Pin Xu ; Rau, Jiann-Yeou. / Using convolutional neural nwtwork for signboard detection on street view images. Paper presented at 39th Asian Conference on Remote Sensing: Remote Sensing Enabling Prosperity, ACRS 2018, Kuala Lumpur, Malaysia.8 p.
@conference{5c03154d78af4067a3cde8e451b24e71,
title = "Using convolutional neural nwtwork for signboard detection on street view images",
abstract = "In order to efficiently build and update store information in digital maps, a convolutional neural network (CNN) model called Faster R-CNN proposed in 2015 is used for signboard detection on street view images. Google's Inception-ResNet-v2 model is the feature extractor in our model and a series of fully-connected layers are used for classification and bounding box regression. In the beginning, a portion of street view images is labelled for training model. Then, the bounding boxes and corresponding probabilities of signboard detection results can be obtained by applying our model to the other portion of street view images. In the evaluation, the precision of our method based on CNN is about 94.87{\%}. In additional evaluations, all the precisions are above 93{\%} after respectively adding Gaussian noise, Gaussian blur, horizontal flip, and change of brightness to the testing images, which shows high potential of our model for future applications. For example, the change analysis or character recognition techniques can be applied to street view images acquired by a mobile mapping system for updating store's attribute as well as geographic location automatically.",
author = "Chen, {Pin Xu} and Jiann-Yeou Rau",
year = "2018",
month = "1",
day = "1",
language = "English",
pages = "1997--2004",
note = "39th Asian Conference on Remote Sensing: Remote Sensing Enabling Prosperity, ACRS 2018 ; Conference date: 15-10-2018 Through 19-10-2018",

}

Chen, PX & Rau, J-Y 2018, 'Using convolutional neural nwtwork for signboard detection on street view images', Paper presented at 39th Asian Conference on Remote Sensing: Remote Sensing Enabling Prosperity, ACRS 2018, Kuala Lumpur, Malaysia, 18-10-15 - 18-10-19 pp. 1997-2004.

Using convolutional neural nwtwork for signboard detection on street view images. / Chen, Pin Xu; Rau, Jiann-Yeou.

2018. 1997-2004 Paper presented at 39th Asian Conference on Remote Sensing: Remote Sensing Enabling Prosperity, ACRS 2018, Kuala Lumpur, Malaysia.

Research output: Contribution to conferencePaper

TY - CONF

T1 - Using convolutional neural nwtwork for signboard detection on street view images

AU - Chen, Pin Xu

AU - Rau, Jiann-Yeou

PY - 2018/1/1

Y1 - 2018/1/1

N2 - In order to efficiently build and update store information in digital maps, a convolutional neural network (CNN) model called Faster R-CNN proposed in 2015 is used for signboard detection on street view images. Google's Inception-ResNet-v2 model is the feature extractor in our model and a series of fully-connected layers are used for classification and bounding box regression. In the beginning, a portion of street view images is labelled for training model. Then, the bounding boxes and corresponding probabilities of signboard detection results can be obtained by applying our model to the other portion of street view images. In the evaluation, the precision of our method based on CNN is about 94.87%. In additional evaluations, all the precisions are above 93% after respectively adding Gaussian noise, Gaussian blur, horizontal flip, and change of brightness to the testing images, which shows high potential of our model for future applications. For example, the change analysis or character recognition techniques can be applied to street view images acquired by a mobile mapping system for updating store's attribute as well as geographic location automatically.

AB - In order to efficiently build and update store information in digital maps, a convolutional neural network (CNN) model called Faster R-CNN proposed in 2015 is used for signboard detection on street view images. Google's Inception-ResNet-v2 model is the feature extractor in our model and a series of fully-connected layers are used for classification and bounding box regression. In the beginning, a portion of street view images is labelled for training model. Then, the bounding boxes and corresponding probabilities of signboard detection results can be obtained by applying our model to the other portion of street view images. In the evaluation, the precision of our method based on CNN is about 94.87%. In additional evaluations, all the precisions are above 93% after respectively adding Gaussian noise, Gaussian blur, horizontal flip, and change of brightness to the testing images, which shows high potential of our model for future applications. For example, the change analysis or character recognition techniques can be applied to street view images acquired by a mobile mapping system for updating store's attribute as well as geographic location automatically.

UR - http://www.scopus.com/inward/record.url?scp=85071875197&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85071875197&partnerID=8YFLogxK

M3 - Paper

AN - SCOPUS:85071875197

SP - 1997

EP - 2004

ER -

Chen PX, Rau J-Y. Using convolutional neural nwtwork for signboard detection on street view images. 2018. Paper presented at 39th Asian Conference on Remote Sensing: Remote Sensing Enabling Prosperity, ACRS 2018, Kuala Lumpur, Malaysia.