Filling the gaps: Reducing the complexity of networks for multi-attribute image aesthetic prediction

Magzhan Kairanbay; John See; Lai-Kuan Wong; Yong-Lian Hii

doi:10.1109/ICIP.2017.8296843

Filling the gaps: Reducing the complexity of networks for multi-attribute image aesthetic prediction

Magzhan Kairanbay, John See, Lai-Kuan Wong, Yong-Lian Hii

School of Mathematical & Computer Sciences

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

8 Citations (Scopus)

Abstract

Computational aesthetics have seen much progress in recent years with the increasing popularity of deep learning methods. In this paper, we present two approaches that leverage on the benefits of using Global Average Pooling (GAP) to reduce the complexity of deep convolutional neural networks. The first model fine-tunes a standard CNN with a newly introduced GAP layer. The second approach extracts global and local CNN codes by reducing the dimensionality of convolution layers with individual GAP operations. We also extend these approaches to a multi-attribute network which uses a style network to regularize the aesthetic network. Experiments demonstrate the capability of attaining comparable accuracy results while reducing training complexity substantially.

Original language	English
Title of host publication	2017 IEEE International Conference on Image Processing (ICIP)
Publisher	IEEE
Pages	3051-3055
Number of pages	5
ISBN (Electronic)	9781509021758
DOIs	https://doi.org/10.1109/ICIP.2017.8296843
Publication status	Published - 22 Feb 2018
Event	24th IEEE International Conference on Image Processing 2017 - Beijing, China Duration: 17 Sept 2017 → 20 Sept 2017

Conference

Conference	24th IEEE International Conference on Image Processing 2017
Abbreviated title	ICIP 2017
Country/Territory	China
City	Beijing
Period	17/09/17 → 20/09/17

Keywords

Aesthetics
Convolutional Neural Network
Global Average Pooling
Multi-attribute Network
Style

ASJC Scopus subject areas

Software
Computer Vision and Pattern Recognition
Signal Processing

Access to Document

10.1109/ICIP.2017.8296843

Cite this

@inproceedings{a35d5e7005b7427a91fa6ecd26b1949c,

title = "Filling the gaps: Reducing the complexity of networks for multi-attribute image aesthetic prediction",

abstract = "Computational aesthetics have seen much progress in recent years with the increasing popularity of deep learning methods. In this paper, we present two approaches that leverage on the benefits of using Global Average Pooling (GAP) to reduce the complexity of deep convolutional neural networks. The first model fine-tunes a standard CNN with a newly introduced GAP layer. The second approach extracts global and local CNN codes by reducing the dimensionality of convolution layers with individual GAP operations. We also extend these approaches to a multi-attribute network which uses a style network to regularize the aesthetic network. Experiments demonstrate the capability of attaining comparable accuracy results while reducing training complexity substantially.",

keywords = "Aesthetics, Convolutional Neural Network, Global Average Pooling, Multi-attribute Network, Style",

author = "Magzhan Kairanbay and John See and Lai-Kuan Wong and Yong-Lian Hii",

note = "Publisher Copyright: {\textcopyright} 2017 IEEE. Copyright: Copyright 2018 Elsevier B.V., All rights reserved.; 24th IEEE International Conference on Image Processing 2017, ICIP 2017 ; Conference date: 17-09-2017 Through 20-09-2017",

year = "2018",

month = feb,

day = "22",

doi = "10.1109/ICIP.2017.8296843",

language = "English",

pages = "3051--3055",

booktitle = "2017 IEEE International Conference on Image Processing (ICIP)",

publisher = "IEEE",

address = "United States",

}

TY - GEN

T1 - Filling the gaps

T2 - 24th IEEE International Conference on Image Processing 2017

AU - Kairanbay, Magzhan

AU - See, John

AU - Wong, Lai-Kuan

AU - Hii, Yong-Lian

PY - 2018/2/22

Y1 - 2018/2/22

N2 - Computational aesthetics have seen much progress in recent years with the increasing popularity of deep learning methods. In this paper, we present two approaches that leverage on the benefits of using Global Average Pooling (GAP) to reduce the complexity of deep convolutional neural networks. The first model fine-tunes a standard CNN with a newly introduced GAP layer. The second approach extracts global and local CNN codes by reducing the dimensionality of convolution layers with individual GAP operations. We also extend these approaches to a multi-attribute network which uses a style network to regularize the aesthetic network. Experiments demonstrate the capability of attaining comparable accuracy results while reducing training complexity substantially.

AB - Computational aesthetics have seen much progress in recent years with the increasing popularity of deep learning methods. In this paper, we present two approaches that leverage on the benefits of using Global Average Pooling (GAP) to reduce the complexity of deep convolutional neural networks. The first model fine-tunes a standard CNN with a newly introduced GAP layer. The second approach extracts global and local CNN codes by reducing the dimensionality of convolution layers with individual GAP operations. We also extend these approaches to a multi-attribute network which uses a style network to regularize the aesthetic network. Experiments demonstrate the capability of attaining comparable accuracy results while reducing training complexity substantially.

KW - Aesthetics

KW - Convolutional Neural Network

KW - Global Average Pooling

KW - Multi-attribute Network

KW - Style

UR - http://www.scopus.com/inward/record.url?scp=85045332958&partnerID=8YFLogxK

U2 - 10.1109/ICIP.2017.8296843

DO - 10.1109/ICIP.2017.8296843

M3 - Conference contribution

AN - SCOPUS:85045332958

SP - 3051

EP - 3055

BT - 2017 IEEE International Conference on Image Processing (ICIP)

PB - IEEE

Y2 - 17 September 2017 through 20 September 2017

ER -

Filling the gaps: Reducing the complexity of networks for multi-attribute image aesthetic prediction

Abstract

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this