CNN-based UGS method using Cartesian-to-polar coordinate transformation

B. S. Kim, J. Y. Sun, S. W. Kim, M. C. Kang, Sung-Jea Ko

Research output: Contribution to journalArticle

Abstract

The main concern of user-guided segmentation (UGS) is to achieve high segmentation accuracy with minimal user interaction. A novel convolutional neural network (CNN)-based UGS method is proposed, which employs a single click as the user interaction. In the proposed method, the input image in the Cartesian coordinate system is first converted into the polar transformed image with the user-guided point (UGP) as the origin of the polar coordinate system. The transformed image not only effectively delivers the UGP to the CNN, but also enables a single-scale convolution kernel to act as a multi-scale kernel, whose receptive field in the Cartesian coordinate system is altered based on the UGP without any extra parameters. In addition, a feature selection module (FSM) is introduced and utilised to additionally extract radial and angular features from the polar transformed image. Experimental results demonstrate that the proposed CNN using the polar transformed image improves the segmentation accuracy (mean intersection over union) by 3.69% on PASCAL VOC 2012 dataset compared with the CNN using the Cartesian coordinate image. The FSM achieves additional performance improvement of 1.32%. Moreover, the proposed method outperforms the conventional non-CNN-based UGS methods by 12.61% on average.

Original languageEnglish
Pages (from-to)1321-1322
Number of pages2
JournalElectronics Letters
Volume54
Issue number23
DOIs
Publication statusPublished - 2018 Nov 15

Fingerprint

Neural networks
Feature extraction
Volatile organic compounds
Convolution

ASJC Scopus subject areas

  • Electrical and Electronic Engineering

Cite this

CNN-based UGS method using Cartesian-to-polar coordinate transformation. / Kim, B. S.; Sun, J. Y.; Kim, S. W.; Kang, M. C.; Ko, Sung-Jea.

In: Electronics Letters, Vol. 54, No. 23, 15.11.2018, p. 1321-1322.

Research output: Contribution to journalArticle

Kim, B. S. ; Sun, J. Y. ; Kim, S. W. ; Kang, M. C. ; Ko, Sung-Jea. / CNN-based UGS method using Cartesian-to-polar coordinate transformation. In: Electronics Letters. 2018 ; Vol. 54, No. 23. pp. 1321-1322.
@article{b5985696f2774cee82d763b8dbb36594,
title = "CNN-based UGS method using Cartesian-to-polar coordinate transformation",
abstract = "The main concern of user-guided segmentation (UGS) is to achieve high segmentation accuracy with minimal user interaction. A novel convolutional neural network (CNN)-based UGS method is proposed, which employs a single click as the user interaction. In the proposed method, the input image in the Cartesian coordinate system is first converted into the polar transformed image with the user-guided point (UGP) as the origin of the polar coordinate system. The transformed image not only effectively delivers the UGP to the CNN, but also enables a single-scale convolution kernel to act as a multi-scale kernel, whose receptive field in the Cartesian coordinate system is altered based on the UGP without any extra parameters. In addition, a feature selection module (FSM) is introduced and utilised to additionally extract radial and angular features from the polar transformed image. Experimental results demonstrate that the proposed CNN using the polar transformed image improves the segmentation accuracy (mean intersection over union) by 3.69{\%} on PASCAL VOC 2012 dataset compared with the CNN using the Cartesian coordinate image. The FSM achieves additional performance improvement of 1.32{\%}. Moreover, the proposed method outperforms the conventional non-CNN-based UGS methods by 12.61{\%} on average.",
author = "Kim, {B. S.} and Sun, {J. Y.} and Kim, {S. W.} and Kang, {M. C.} and Sung-Jea Ko",
year = "2018",
month = "11",
day = "15",
doi = "10.1049/el.2018.5051",
language = "English",
volume = "54",
pages = "1321--1322",
journal = "Electronics Letters",
issn = "0013-5194",
publisher = "Institution of Engineering and Technology",
number = "23",

}

TY - JOUR

T1 - CNN-based UGS method using Cartesian-to-polar coordinate transformation

AU - Kim, B. S.

AU - Sun, J. Y.

AU - Kim, S. W.

AU - Kang, M. C.

AU - Ko, Sung-Jea

PY - 2018/11/15

Y1 - 2018/11/15

N2 - The main concern of user-guided segmentation (UGS) is to achieve high segmentation accuracy with minimal user interaction. A novel convolutional neural network (CNN)-based UGS method is proposed, which employs a single click as the user interaction. In the proposed method, the input image in the Cartesian coordinate system is first converted into the polar transformed image with the user-guided point (UGP) as the origin of the polar coordinate system. The transformed image not only effectively delivers the UGP to the CNN, but also enables a single-scale convolution kernel to act as a multi-scale kernel, whose receptive field in the Cartesian coordinate system is altered based on the UGP without any extra parameters. In addition, a feature selection module (FSM) is introduced and utilised to additionally extract radial and angular features from the polar transformed image. Experimental results demonstrate that the proposed CNN using the polar transformed image improves the segmentation accuracy (mean intersection over union) by 3.69% on PASCAL VOC 2012 dataset compared with the CNN using the Cartesian coordinate image. The FSM achieves additional performance improvement of 1.32%. Moreover, the proposed method outperforms the conventional non-CNN-based UGS methods by 12.61% on average.

AB - The main concern of user-guided segmentation (UGS) is to achieve high segmentation accuracy with minimal user interaction. A novel convolutional neural network (CNN)-based UGS method is proposed, which employs a single click as the user interaction. In the proposed method, the input image in the Cartesian coordinate system is first converted into the polar transformed image with the user-guided point (UGP) as the origin of the polar coordinate system. The transformed image not only effectively delivers the UGP to the CNN, but also enables a single-scale convolution kernel to act as a multi-scale kernel, whose receptive field in the Cartesian coordinate system is altered based on the UGP without any extra parameters. In addition, a feature selection module (FSM) is introduced and utilised to additionally extract radial and angular features from the polar transformed image. Experimental results demonstrate that the proposed CNN using the polar transformed image improves the segmentation accuracy (mean intersection over union) by 3.69% on PASCAL VOC 2012 dataset compared with the CNN using the Cartesian coordinate image. The FSM achieves additional performance improvement of 1.32%. Moreover, the proposed method outperforms the conventional non-CNN-based UGS methods by 12.61% on average.

UR - http://www.scopus.com/inward/record.url?scp=85056908769&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85056908769&partnerID=8YFLogxK

U2 - 10.1049/el.2018.5051

DO - 10.1049/el.2018.5051

M3 - Article

AN - SCOPUS:85056908769

VL - 54

SP - 1321

EP - 1322

JO - Electronics Letters

JF - Electronics Letters

SN - 0013-5194

IS - 23

ER -