Sequential gallery for interactive visual design optimization

Abstract

Visual design tasks often involve tuning many design parameters. For example, color grading of a photograph involves many parameters, some of which non-expert users might be unfamiliar with. We propose a novel user-in-the-loop optimization method that allows users to efficiently find an appropriate parameter set by exploring such a high-dimensional design space through much easier two-dimensional search subtasks. This method, called sequential plane search , is based on Bayesian optimization to keep necessary queries to users as few as possible. To help users respond to plane-search queries, we also propose using a gallery-based interface that provides options in the two-dimensional subspace arranged in an adaptive grid view. We call this interactive framework Sequential Gallery since users sequentially select the best option from the options provided by the interface. Our experiment with synthetic functions shows that our sequential plane search can find satisfactory solutions in fewer iterations than baselines. We also conducted a preliminary user study, results of which suggest that novices can effectively complete search tasks with Sequential Gallery in a photo-enhancement scenario.

Generative Design Inspiration for Glyphs with Diatoms

“…This automation is well-suited for drawing and photo editing applications [45,52,77], where an interactive parameter space can quickly produce many variants. Interactive design externalization can facilitate browsing and presenting design alternatives [14] as well as a sequential exploration of a parameter space [43], either by individuals or by teams of designers engaged in a process of collaborative critique [54]. Externalization for visualization design.…”

Pareto gamuts

“…Multiobjective optimization algorithms are classified into three categories, based on the timing of user guidance relative to optimization: before (a priori), during (interactive), or after (a posteriori) [Van Veldhuizen and Lamont 2000]. Interactive and a priori methods allow users to guide the optimization toward "desirable" regions [Koyama et al 2020;Marler and Arora 2004;Meignan et al 2015;Ruiz et al 2019]. These approaches require relatively few samples, but user-guided exploration is biased toward known solutions, which reinforces user expectations and impedes novel insights.…”

Differentiable 3D CAD Programs for Bidirectional Editing

“…This approach is not automatic, and when a given user intent is not within the program's output space, many objectives tend to converge to similar results while still differing from a user's intent. To address this, future work can use existing approaches to refining discrete parameter spaces, including design galleries [Marks et al 1997;Shimizu et al 2020] and user-in-the loop optimization [Koyama et al 2020], in combination with program synthesis as a possible avenue for exploration. We suggest an approach where users can select an area of interest and a system can synthesize possible parameterizations that contain the desired output, which can then be explored using bidirectional editing.…”