Finding image features associated with high aesthetic value by machine learning

Authors:
Vic Ciesielski;Perry Barile;Karen Trist
Affiliations:
School of Computer Science and Information Technology, RMIT University, Melbourne, Vic, Australia;School of Computer Science and Information Technology, RMIT University, Melbourne, Vic, Australia;School of Media and Communication, RMIT University, Melbourne, Vic, Australia
Venue:
EvoMUSART'13 Proceedings of the Second international conference on Evolutionary and Biologically Inspired Music, Sound, Art and Design
Year:
2013

Citing 4
Cited 0

Data mining: practical machine learning tools and techniques with Java implementations

Data mining: practical machine learning tools and techniques with Java implementations
Evolving images for entertainment

IE '07 Proceedings of the 4th Australasian conference on Interactive entertainment
Studying aesthetics in photographic images using a computational approach

ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part III
Open problems in evolutionary music and art

EC'05 Proceedings of the 3rd European conference on Applications of Evolutionary Computing

Quantified Score

Hi-index	0.00

Visualization

Abstract

A major goal of evolutionary art is to get images of high aesthetic value. We assume that some features of images are associated with high aesthetic value and want to find them. We have taken two image databases that have been rated by humans, a photographic database and one of abstract images generated by evolutionary art software. We have computed 55 features for each database. We have extracted two categories of rankings, the lowest and the highest. Using feature extraction methods from machine learning we have identified the features most associated with differences. For the photographic images the key features are wavelet and texture features. For the abstract images the features are colour based features.