Deep correlation features for image style classification

Wei Ta Chu; Yi Ling Wu

doi:10.1145/2964284.2967251

Deep correlation features for image style classification

Wei Ta Chu, Yi Ling Wu

Department of Computer Science and Information Engineering

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

23 Citations (Scopus)

Abstract

This paper presents a comprehensive study of deep correlation features on image style classification. Inspired by that correlation between feature maps can effectively describe image texture, we design and transform various such correlations into style vectors, and investigate classification performance brought by different variants. In addition to intra-layer correlation, we also propose inter-layer correlation and verify its benefit. Through extensive experiments on image style classification and artist classification, we demonstrate that the proposed style vectors significantly outperforms CNN features coming from fully-connected layers, as well as outperforms the state-of-the-art deep representation.

Original language	English
Title of host publication	MM 2016 - Proceedings of the 2016 ACM Multimedia Conference
Publisher	Association for Computing Machinery, Inc
Pages	402-406
Number of pages	5
ISBN (Electronic)	9781450336031
DOIs	https://doi.org/10.1145/2964284.2967251
Publication status	Published - 2016 Oct 1
Event	24th ACM Multimedia Conference, MM 2016 - Amsterdam, United Kingdom Duration: 2016 Oct 15 → 2016 Oct 19

Publication series

Name	MM 2016 - Proceedings of the 2016 ACM Multimedia Conference

Conference

Conference	24th ACM Multimedia Conference, MM 2016
Country	United Kingdom
City	Amsterdam
Period	16-10-15 → 16-10-19

All Science Journal Classification (ASJC) codes

Computer Graphics and Computer-Aided Design
Human-Computer Interaction
Computer Vision and Pattern Recognition
Software

Access to Document

10.1145/2964284.2967251

Fingerprint Dive into the research topics of 'Deep correlation features for image style classification'. Together they form a unique fingerprint.

Cite this

@inproceedings{dba46e67c3fa4714810655baf70c312a,

title = "Deep correlation features for image style classification",

abstract = "This paper presents a comprehensive study of deep correlation features on image style classification. Inspired by that correlation between feature maps can effectively describe image texture, we design and transform various such correlations into style vectors, and investigate classification performance brought by different variants. In addition to intra-layer correlation, we also propose inter-layer correlation and verify its benefit. Through extensive experiments on image style classification and artist classification, we demonstrate that the proposed style vectors significantly outperforms CNN features coming from fully-connected layers, as well as outperforms the state-of-the-art deep representation.",

author = "Chu, {Wei Ta} and Wu, {Yi Ling}",

year = "2016",

month = oct,

day = "1",

doi = "10.1145/2964284.2967251",

language = "English",

series = "MM 2016 - Proceedings of the 2016 ACM Multimedia Conference",

publisher = "Association for Computing Machinery, Inc",

pages = "402--406",

booktitle = "MM 2016 - Proceedings of the 2016 ACM Multimedia Conference",

note = "24th ACM Multimedia Conference, MM 2016 ; Conference date: 15-10-2016 Through 19-10-2016",

}

Chu, WT & Wu, YL 2016, Deep correlation features for image style classification. in MM 2016 - Proceedings of the 2016 ACM Multimedia Conference. MM 2016 - Proceedings of the 2016 ACM Multimedia Conference, Association for Computing Machinery, Inc, pp. 402-406, 24th ACM Multimedia Conference, MM 2016, Amsterdam, United Kingdom, 16-10-15. https://doi.org/10.1145/2964284.2967251

TY - GEN

T1 - Deep correlation features for image style classification

AU - Chu, Wei Ta

AU - Wu, Yi Ling

PY - 2016/10/1

Y1 - 2016/10/1

N2 - This paper presents a comprehensive study of deep correlation features on image style classification. Inspired by that correlation between feature maps can effectively describe image texture, we design and transform various such correlations into style vectors, and investigate classification performance brought by different variants. In addition to intra-layer correlation, we also propose inter-layer correlation and verify its benefit. Through extensive experiments on image style classification and artist classification, we demonstrate that the proposed style vectors significantly outperforms CNN features coming from fully-connected layers, as well as outperforms the state-of-the-art deep representation.

AB - This paper presents a comprehensive study of deep correlation features on image style classification. Inspired by that correlation between feature maps can effectively describe image texture, we design and transform various such correlations into style vectors, and investigate classification performance brought by different variants. In addition to intra-layer correlation, we also propose inter-layer correlation and verify its benefit. Through extensive experiments on image style classification and artist classification, we demonstrate that the proposed style vectors significantly outperforms CNN features coming from fully-connected layers, as well as outperforms the state-of-the-art deep representation.

UR - http://www.scopus.com/inward/record.url?scp=84994681110&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84994681110&partnerID=8YFLogxK

U2 - 10.1145/2964284.2967251

DO - 10.1145/2964284.2967251

M3 - Conference contribution

AN - SCOPUS:84994681110

T3 - MM 2016 - Proceedings of the 2016 ACM Multimedia Conference

SP - 402

EP - 406

BT - MM 2016 - Proceedings of the 2016 ACM Multimedia Conference

PB - Association for Computing Machinery, Inc

T2 - 24th ACM Multimedia Conference, MM 2016

Y2 - 15 October 2016 through 19 October 2016

ER -