19
Evaluating Art by Measuring Complexity 18 November 2014 Eelco den Heijer

Evaluating Art by measuring Complexity

Embed Size (px)

Citation preview

Page 1: Evaluating Art by measuring Complexity

Evaluating Art by MeasuringComplexity

18 November 2014

Eelco den Heijer

Page 2: Evaluating Art by measuring Complexity

Introduction

• Autonomous Evolutionary Art

• Evolving art, images; using an aesthetic measure as the fitness function

• Complexity of the image is (only) one aspect of aesthetic appeal

• Relation between Art and Complexity is unclear

Page 3: Evaluating Art by measuring Complexity
Page 4: Evaluating Art by measuring Complexity
Page 5: Evaluating Art by measuring Complexity
Page 6: Evaluating Art by measuring Complexity

Aesthetic Measures

• Compute aspects of beauty

• Information Complexity, Image Complexity, Processing Complexity

• Colour Harmony

• Symmetry & Balance

• Colour gradients

• Contrast

Page 7: Evaluating Art by measuring Complexity

Shannon Entropy

• Calculate entropy of the intensity of the pixels of an image

• Liveliness of an image

• Rigau et al, 2008 - Informational aesthetics measures. IEEE Computer Graphics and Applications, 28(2):24–34, 2008

Page 8: Evaluating Art by measuring Complexity

Fractal Dimension

• Approximate self-similarity

• Use box-counting method

• Very difficult to ‘satisfy’ with our representation & EC parameters

Page 9: Evaluating Art by measuring Complexity

Machado/ Cardoso

• Calculate Image Complexity (IC) and Processing Complexity (PC)

• “Images that are visually complex, but are processed easily have the highest aesthetic value”

• Penousal Machado and Amílcar Cardoso. Computing aesthetics. In Proceedings of the Brazilian Symposium on Artificial Intelligence, SBIA-98, pages 219–229. Springer- Verlag, 1998.

Page 10: Evaluating Art by measuring Complexity

Machado/ Cardoso (2)

• Image complexity

• RMS(I) = difference in pixels between original image and compressed image (how well can you compress an image)

• Uses JPEG compressor, 75% quality setting

Page 11: Evaluating Art by measuring Complexity

Machado/ Cardoso (3)• Processing Complexity

• Calculates complexity at ‘multiple time points’

• Divide image into 4 equal parts, calculate processing complexity for each part

• Estimate the processing complexity using

• Fractal compression (Machado & Cardoso)

• JPEG 2000 compression (den Heijer & Eiben)

• Run-Length Encoding (Atkins et al)

Page 12: Evaluating Art by measuring Complexity

Facticity

• “Entropy and Kolmogorov complexity do not necessarily measure the interestingness of a system of a data set”

• Describes the amount of meaningful information of a dataset

• Pieter W. Adriaans. Between order and chaos: The quest for meaningful information. Theory Comput. Syst., 45(4):650–674, 2009

Page 13: Evaluating Art by measuring Complexity
Page 14: Evaluating Art by measuring Complexity
Page 15: Evaluating Art by measuring Complexity
Page 16: Evaluating Art by measuring Complexity
Page 17: Evaluating Art by measuring Complexity
Page 18: Evaluating Art by measuring Complexity

Discussion

• The use of simple complexity estimation tools is merely the beginning in calculating aesthetic appreciation

• How does the brain process complexity?

• And how is this linked to aesthetic value?

• Processing fluency theory (Reber, 2004)

• We probably need better models of human visual processing

Page 19: Evaluating Art by measuring Complexity

Questions?

[email protected]