Superpixel convolution for segmentation

Teppei Suzuki, Shuichi Akizuki, Naoki Kato, Yoshimitsu Aoki

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this paper, we propose a novel segmentation algorithm based on convolutional neural networks (CNNs) on superpix-else CNNs are powerful methods for several computer vision tasks, but spatial information disappears through the pooling process. Moreover, since pooling compresses different types of pixels into a single value, pooling sometimes negatively affects the results of inference in segmentation task. We use superpixel pooling instead of general pooling to resolve this problem. However, general CNNs can't use superpixel images in which the adjacency relationships between pixels are broken. Therefore, we define CNNs and Dilated Convolution on superpixels. Finally, we show the effectiveness of proposed method on an HKU-IS dataset.

Original languageEnglish
Title of host publication2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings
PublisherIEEE Computer Society
Pages3249-3253
Number of pages5
ISBN (Electronic)9781479970612
DOIs
Publication statusPublished - 2018 Aug 29
Event25th IEEE International Conference on Image Processing, ICIP 2018 - Athens, Greece
Duration: 2018 Oct 72018 Oct 10

Publication series

NameProceedings - International Conference on Image Processing, ICIP
ISSN (Print)1522-4880

Conference

Conference25th IEEE International Conference on Image Processing, ICIP 2018
CountryGreece
CityAthens
Period18/10/718/10/10

Fingerprint

Convolution
Neural networks
Pixels
Computer vision

Keywords

  • Convolutional Neural Networks
  • Saliency Object Detection
  • Segmentation
  • Super-pixel

ASJC Scopus subject areas

  • Software
  • Computer Vision and Pattern Recognition
  • Signal Processing

Cite this

Suzuki, T., Akizuki, S., Kato, N., & Aoki, Y. (2018). Superpixel convolution for segmentation. In 2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings (pp. 3249-3253). [8451721] (Proceedings - International Conference on Image Processing, ICIP). IEEE Computer Society. https://doi.org/10.1109/ICIP.2018.8451721

Superpixel convolution for segmentation. / Suzuki, Teppei; Akizuki, Shuichi; Kato, Naoki; Aoki, Yoshimitsu.

2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings. IEEE Computer Society, 2018. p. 3249-3253 8451721 (Proceedings - International Conference on Image Processing, ICIP).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Suzuki, T, Akizuki, S, Kato, N & Aoki, Y 2018, Superpixel convolution for segmentation. in 2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings., 8451721, Proceedings - International Conference on Image Processing, ICIP, IEEE Computer Society, pp. 3249-3253, 25th IEEE International Conference on Image Processing, ICIP 2018, Athens, Greece, 18/10/7. https://doi.org/10.1109/ICIP.2018.8451721
Suzuki T, Akizuki S, Kato N, Aoki Y. Superpixel convolution for segmentation. In 2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings. IEEE Computer Society. 2018. p. 3249-3253. 8451721. (Proceedings - International Conference on Image Processing, ICIP). https://doi.org/10.1109/ICIP.2018.8451721
Suzuki, Teppei ; Akizuki, Shuichi ; Kato, Naoki ; Aoki, Yoshimitsu. / Superpixel convolution for segmentation. 2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings. IEEE Computer Society, 2018. pp. 3249-3253 (Proceedings - International Conference on Image Processing, ICIP).
@inproceedings{3496e341e74645deb2fc2e295a81f92e,
title = "Superpixel convolution for segmentation",
abstract = "In this paper, we propose a novel segmentation algorithm based on convolutional neural networks (CNNs) on superpix-else CNNs are powerful methods for several computer vision tasks, but spatial information disappears through the pooling process. Moreover, since pooling compresses different types of pixels into a single value, pooling sometimes negatively affects the results of inference in segmentation task. We use superpixel pooling instead of general pooling to resolve this problem. However, general CNNs can't use superpixel images in which the adjacency relationships between pixels are broken. Therefore, we define CNNs and Dilated Convolution on superpixels. Finally, we show the effectiveness of proposed method on an HKU-IS dataset.",
keywords = "Convolutional Neural Networks, Saliency Object Detection, Segmentation, Super-pixel",
author = "Teppei Suzuki and Shuichi Akizuki and Naoki Kato and Yoshimitsu Aoki",
year = "2018",
month = "8",
day = "29",
doi = "10.1109/ICIP.2018.8451721",
language = "English",
series = "Proceedings - International Conference on Image Processing, ICIP",
publisher = "IEEE Computer Society",
pages = "3249--3253",
booktitle = "2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings",
address = "United States",

}

TY - GEN

T1 - Superpixel convolution for segmentation

AU - Suzuki, Teppei

AU - Akizuki, Shuichi

AU - Kato, Naoki

AU - Aoki, Yoshimitsu

PY - 2018/8/29

Y1 - 2018/8/29

N2 - In this paper, we propose a novel segmentation algorithm based on convolutional neural networks (CNNs) on superpix-else CNNs are powerful methods for several computer vision tasks, but spatial information disappears through the pooling process. Moreover, since pooling compresses different types of pixels into a single value, pooling sometimes negatively affects the results of inference in segmentation task. We use superpixel pooling instead of general pooling to resolve this problem. However, general CNNs can't use superpixel images in which the adjacency relationships between pixels are broken. Therefore, we define CNNs and Dilated Convolution on superpixels. Finally, we show the effectiveness of proposed method on an HKU-IS dataset.

AB - In this paper, we propose a novel segmentation algorithm based on convolutional neural networks (CNNs) on superpix-else CNNs are powerful methods for several computer vision tasks, but spatial information disappears through the pooling process. Moreover, since pooling compresses different types of pixels into a single value, pooling sometimes negatively affects the results of inference in segmentation task. We use superpixel pooling instead of general pooling to resolve this problem. However, general CNNs can't use superpixel images in which the adjacency relationships between pixels are broken. Therefore, we define CNNs and Dilated Convolution on superpixels. Finally, we show the effectiveness of proposed method on an HKU-IS dataset.

KW - Convolutional Neural Networks

KW - Saliency Object Detection

KW - Segmentation

KW - Super-pixel

UR - http://www.scopus.com/inward/record.url?scp=85062894316&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85062894316&partnerID=8YFLogxK

U2 - 10.1109/ICIP.2018.8451721

DO - 10.1109/ICIP.2018.8451721

M3 - Conference contribution

AN - SCOPUS:85062894316

T3 - Proceedings - International Conference on Image Processing, ICIP

SP - 3249

EP - 3253

BT - 2018 IEEE International Conference on Image Processing, ICIP 2018 - Proceedings

PB - IEEE Computer Society

ER -