We are increasingly producing and consuming digitally created or mediated visual content. Computational models of visual aesthetics could potentially transform the way in which humans interact with computers and navigate through available visual content. They are challenging to develop due to the lack of functional and interpretable metrics, noise in naturally occurring data sets, and unreliable self-reported data due to the different interpretations of embedded content. Early work in psychology on art and visual perception (Arnheim’74) identified concepts like symmetry, rhythm, contrast, etc. that correspond to cross-domain features of visual aesthetics. In this talk, I will present a process for modeling aesthetic evaluation of visual artifacts or experiences through preference elicitation on synthetic data across domains. Specifically, I will show results in the domains of photographic composition and full-body gestural performance (dance). I will then discuss issues related to the design of experiments for data collection, feature modeling and selection, and applications of machine learning algorithms for learning preferences. Finally, I will conclude with a vision for how this work contributes to the broader goal of developing intelligent systems with the ability to interpret and create visual narratives.

Arnav Jhala is an Associate Professor of Computational Media at the University of California, Santa Cruz. His research interests lie at the intersection of artificial intelligence and digital media, particularly in the areas of computer games, visual communication, and narrative intelligence. He heads the Computational Cinematics Studio at UC Santa Cruz which has been supported through generous funding from the Baskin School of Engineering, NSF, DARPA, EU FP7, Microsoft, Facebook, and EBay. Arnav came to Santa Cruz from IT University of Copenhagen, where he was an assistant professor in the Center for Computer Games Research. As an intern at the University of Southern California's Institute for Creative Technologies and Paramount Pictures he was part of the Leaders project, applying Hollywood storytelling techniques to the design of story-based training applications with branching storylines. He also spent a summer working on the America's Army: Adaptive Thinking and Leadership game at Virtual Heroes, Inc., a leading serious games developer. Arnav holds Ph.D. and M.S. degrees in Computer Science from the North Carolina State University and B.Eng. in Computer Engineering from Gujarat University.