Predicting the aesthetics of dynamic generative artwork based on statistical image features: A time-dependent model.

Area of Science:

• Computer Science
• Art and Aesthetics
• Human-Computer Interaction

Background:

• Automated aesthetic assessment models primarily focus on static visual art.
• Dynamic visual art, particularly computer-based forms, offers unique artistic expression and audience engagement.
• A dedicated model for dynamic art evaluation is needed to guide artists and enhance creative processes.

Purpose of the Study:

• To develop a time-dependent model for predicting the aesthetic appeal of dynamic generative artworks.
• To identify key visual features that influence the aesthetic quality of evolving digital art.
• To provide a framework for evaluating and improving dynamic generative art.

Main Methods:

• Eight generative artworks were created using a common method.
• A time-dependent model was established to predict aesthetics based on visual features.
• Panel regression analysis was employed to assess the impact of time-varying features on aesthetic appeal.

Main Results:

• Selected image features effectively captured the evolving characteristics of the dynamic artworks.
• Aesthetic appeal was significantly influenced by luminance distribution skewness, vertical symmetry, and mean hue value.
• The study demonstrated that integrating temporal image features predicts the aesthetic appeal of dynamic generative art.

Conclusions:

• A novel approach for evaluating dynamic generative art aesthetics has been established.
• Specific visual features, analyzed over time, are crucial predictors of aesthetic appeal in digital art.
• This research contributes to the field of automated art assessment by addressing the unique challenges of dynamic visual forms.