Quantifying Aesthetics of Visual Design Applied to Automatic Design (eBook)

Dr. Ali Jahanian graduated with a Ph.D. from the School of Electrical and Computer Engineering at Purdue University, and is currently at the Massachusetts Institute of Technology Computer Science and Artificial Intelligence Laboratory.

Preface by Supervisor 8
Acknowledgments 10
Contents 12
1 Introduction 16
1.1 Introduction 16
1.2 Motivation 16
1.3 Thesis Contributions 17
1.3.1 Identifying Challenges 17
1.3.2 Defining Research Strategies 18
1.3.3 Establishing a Taxonomy 19
1.3.4 Compiling Design Guidelines 19
1.3.5 Automating Design Processes 19
1.3.6 Semantic Design Mining 20
1.3.7 Quantitatively Revisiting Two Design Theories via Large-Scale Data 20
1.3.8 Deploying Crowdsourcing for Design 20
1.3.9 Devising Recommendation Systems for Design 21
Bibliography 21
2 On the Legitimacy of Quantifying Aesthetics 22
2.1 Theoretical Considerations 22
2.2 Taxonomy 24
2.2.1 Automatic Approaches 25
2.2.2 Human Inspection Approaches 26
2.3 Conclusion 26
Bibliography 27
3 Design Mining Color Semantics 30
3.1 Introduction 30
3.2 Theory 32
3.2.1 Color Cognition 33
3.2.2 Color Naming 33
3.2.3 Color Meanings and Semantics 33
3.2.4 Color Semantics, Emotions, and Preferences 34
3.2.5 Color Semantics and Cross-Cultural Considerations 35
3.2.6 Color Semantics in Applied Arts 35
3.2.7 Color Semantics in HCI 35
3.3 Related Work 36
3.3.1 Color Semantics and Meanings 36
3.3.2 LDA Topic Modeling 37
3.3.3 User Studies via Crowdsourcing 37
3.3.4 Click Modeling 38
3.4 Data Collection 38
3.4.1 Preprocessing 39
3.4.2 Word Vocabulary 39
3.5 Statistical Model 39
3.5.1 Review of LDA-Dual Model for Color Semantics 41
3.5.2 Inference 43
3.5.3 Effect of Color Basis 46
3.6 Interpreting Model Output 46
3.6.1 From Color Histograms to Color Palettes 47
3.6.2 From Weighted Bag of Words to Word Clouds 48
3.7 User Study 50
3.7.1 Formative Study 50
3.7.2 Summative Study 51
3.7.2.1 Participants 51
3.7.2.2 Stimuli and Procedure 51
3.8 Interpreting the User Study 52
3.8.1 Statistical Model 52
3.8.2 Analyzing the Results 54
3.8.3 Association Directionality 56
3.9 Applications 56
3.9.1 Color Palette Selection Using Semantics 57
3.9.2 Design Example Recommendation 58
3.9.3 Image Retrieval Using Color Semantics 59
3.9.4 Image Color Selection Using Semantics 59
3.9.5 Image Recoloring Using Semantics 61
3.10 Conclusion and Future Work 61
3.10.1 Visual Design Language and User Interaction 62
3.10.2 Quantifying Aesthetics 63
Bibliography 63
4 Design Mining Visual Balance 71
4.1 Introduction 71
4.2 Theory 72
4.2.1 Visual Balance in Spatial Composition 72
4.2.2 Visual Rightness 73
4.3 Modeling Framework 74
4.3.1 EM for Gaussian Mixtures 77
4.3.2 Dataset 77
4.4 Results 77
4.5 Discussion and Future Work 78
Bibliography 79
5 Automatic Design of Self-Published Media: A Case Study of Magazine Covers 83
5.1 Introduction 83
5.2 Theoretical Considerations 84
5.2.1 Principles of Design 85
5.2.1.1 Unity 85
5.2.1.2 Contrast 85
5.2.1.3 Rhythm 86
5.2.1.4 Balance 86
5.2.2 Elements of Design 86
5.2.2.1 Cover Image 86
5.2.2.2 Masthead 87
5.2.2.3 Cover Lines 87
5.2.2.4 Price, Date, and Bar Code 88
5.2.2.5 Spine 88
5.2.2.6 Back Cover 88
5.3 Related Work 88
5.3.1 Automated Layout 88
5.3.2 Color Design 89
5.3.3 Typography 89
5.3.4 Computational Aesthetics 90
5.4 Automatic Design System 90
5.4.1 Software Framework Overview 91
5.5 Layout 92
5.5.1 Defining Space 92
5.5.2 Visual Balance Considerations 93
5.6 Typography 96
5.6.1 Masthead 97
5.6.2 Headline and Byline 98
5.6.3 Cover Lines 98
5.7 Design of Color 100
5.7.1 Masthead Color 101
5.7.2 Cover Lines Colors 103
5.7.3 Color Aesthetics and Text Legibility 103
5.8 Experimental Results 105
5.8.1 Results 105
5.9 Conclusion and Future Work 110
Bibliography 110
6 Recommendation System for Automatic Design 114
6.1 Introduction 114
6.2 Scenario 116
6.3 Related Work 116
6.4 Software Framework 117
6.4.1 Schematic View 117
6.5 Input User Interface 118
6.6 Evaluation of Input Photos 119
6.7 Design User Interface 123
6.8 Personalization of Designs 123
6.9 Experimental Results 125
6.10 Conclusion and Future Work 126
Bibliography 127
A Design Mining Color Semantics 129
Bibliography 137
B Derivations of Gibbs Sampling for LDA-Dual 138
B.1 Notation 138
B.2 Joint Probability 139
B.3 Integrating Out Multinomials 140
Bibliography 144
Index 145