Randomly shuffled convolution for self-supervised representation learning

Youngjin Oh; Minkyu Jeon; Dohwan Ko; Hyunwoo J. Kim

doi:10.1016/j.ins.2022.11.022

Randomly shuffled convolution for self-supervised representation learning

Youngjin Oh, Minkyu Jeon, Dohwan Ko, Hyunwoo J. Kim

Research output: Contribution to journal › Article › peer-review

1 Citation (Scopus)

Abstract

Many self-supervised representation learning methods have achieved high performance in image classification tasks. However, these methods have limited performance on localization tasks such as object detection or semantic segmentation. Most self-supervised representation learning methods are optimized with only one global representation, which does not pay much attention to the spatial information in an image. We propose a simple and effective method that uses the positional relationships between the entities in an image by shuffling the convolution kernels. Our method extends current self-supervised learning and calculates the pixel-wise (dis) similarities between the output of the standard convolution kernels and that of the randomly shuffled convolution kernels. Our proposed method achieves higher performance on object detection, instance segmentation, and semantic segmentation when attached to recent self-supervised learning methods.

Original language	English
Pages (from-to)	206-219
Number of pages	14
Journal	Information Sciences
Volume	623
DOIs	https://doi.org/10.1016/j.ins.2022.11.022
Publication status	Published - 2023 Apr

Bibliographical note

Funding Information:
This study was partly supported by a grant from the Korea Health Technology R&D Project, through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health and Welfare, Republic of Korea (Grant No., HR20C0021); the ICT Creative Consilience program (IITP-2022-2020-0-01819) supervised by the IITP; and an IITP grant funded by the Korea government (MSIT) (No.2019–0-00533, Research on CPU vulnerability detection and validation).

Publisher Copyright:
© 2022 The Author(s)

Keywords

Deep learning
Machine learning
Representation learning
Self-supervised learning
Unsupervised learning

ASJC Scopus subject areas

Software
Control and Systems Engineering
Theoretical Computer Science
Computer Science Applications
Information Systems and Management
Artificial Intelligence

Access to Document

10.1016/j.ins.2022.11.022

Cite this

@article{d9ed9a870d6b4b37b1dae888443428a5,

title = "Randomly shuffled convolution for self-supervised representation learning",

abstract = "Many self-supervised representation learning methods have achieved high performance in image classification tasks. However, these methods have limited performance on localization tasks such as object detection or semantic segmentation. Most self-supervised representation learning methods are optimized with only one global representation, which does not pay much attention to the spatial information in an image. We propose a simple and effective method that uses the positional relationships between the entities in an image by shuffling the convolution kernels. Our method extends current self-supervised learning and calculates the pixel-wise (dis) similarities between the output of the standard convolution kernels and that of the randomly shuffled convolution kernels. Our proposed method achieves higher performance on object detection, instance segmentation, and semantic segmentation when attached to recent self-supervised learning methods.",

keywords = "Deep learning, Machine learning, Representation learning, Self-supervised learning, Unsupervised learning",

author = "Youngjin Oh and Minkyu Jeon and Dohwan Ko and Kim, {Hyunwoo J.}",

note = "Funding Information: This study was partly supported by a grant from the Korea Health Technology R&D Project, through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health and Welfare, Republic of Korea (Grant No., HR20C0021); the ICT Creative Consilience program (IITP-2022-2020-0-01819) supervised by the IITP; and an IITP grant funded by the Korea government (MSIT) (No.2019–0-00533, Research on CPU vulnerability detection and validation). Publisher Copyright: {\textcopyright} 2022 The Author(s)",

year = "2023",

month = apr,

doi = "10.1016/j.ins.2022.11.022",

language = "English",

volume = "623",

pages = "206--219",

journal = "Information Sciences",

issn = "0020-0255",

publisher = "Elsevier Inc.",

}

TY - JOUR

T1 - Randomly shuffled convolution for self-supervised representation learning

AU - Oh, Youngjin

AU - Jeon, Minkyu

AU - Ko, Dohwan

AU - Kim, Hyunwoo J.

N1 - Funding Information: This study was partly supported by a grant from the Korea Health Technology R&D Project, through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health and Welfare, Republic of Korea (Grant No., HR20C0021); the ICT Creative Consilience program (IITP-2022-2020-0-01819) supervised by the IITP; and an IITP grant funded by the Korea government (MSIT) (No.2019–0-00533, Research on CPU vulnerability detection and validation). Publisher Copyright: © 2022 The Author(s)

PY - 2023/4

Y1 - 2023/4

N2 - Many self-supervised representation learning methods have achieved high performance in image classification tasks. However, these methods have limited performance on localization tasks such as object detection or semantic segmentation. Most self-supervised representation learning methods are optimized with only one global representation, which does not pay much attention to the spatial information in an image. We propose a simple and effective method that uses the positional relationships between the entities in an image by shuffling the convolution kernels. Our method extends current self-supervised learning and calculates the pixel-wise (dis) similarities between the output of the standard convolution kernels and that of the randomly shuffled convolution kernels. Our proposed method achieves higher performance on object detection, instance segmentation, and semantic segmentation when attached to recent self-supervised learning methods.

AB - Many self-supervised representation learning methods have achieved high performance in image classification tasks. However, these methods have limited performance on localization tasks such as object detection or semantic segmentation. Most self-supervised representation learning methods are optimized with only one global representation, which does not pay much attention to the spatial information in an image. We propose a simple and effective method that uses the positional relationships between the entities in an image by shuffling the convolution kernels. Our method extends current self-supervised learning and calculates the pixel-wise (dis) similarities between the output of the standard convolution kernels and that of the randomly shuffled convolution kernels. Our proposed method achieves higher performance on object detection, instance segmentation, and semantic segmentation when attached to recent self-supervised learning methods.

KW - Deep learning

KW - Machine learning

KW - Representation learning

KW - Self-supervised learning

KW - Unsupervised learning

UR - http://www.scopus.com/inward/record.url?scp=85144608413&partnerID=8YFLogxK

U2 - 10.1016/j.ins.2022.11.022

DO - 10.1016/j.ins.2022.11.022

M3 - Article

AN - SCOPUS:85144608413

SN - 0020-0255

VL - 623

SP - 206

EP - 219

JO - Information Sciences

JF - Information Sciences

ER -

Randomly shuffled convolution for self-supervised representation learning

Abstract

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this