Retrieve-Then-Adapt: Example-based Automatic Generation for Proportion-related Infographics

Infographic is a data visualization technique which combines graphic and textual descriptions in an aesthetic effective manner. Creating infographics difficult time-consuming process often requires significant attempts adjustments even for experienced designers, not to mention novice users with limited design expertise. Recently, few approaches have been proposed automate the creation by applying predefined blueprints user information. However, are hard create, hence volume diversity. In contrast, good infogrpahics created professionals accumulated on Internet rapidly. These online examples represent wide variety of styles, serve as exemplars or inspiration people who like create their own infographics. Based these observations, we propose generate automatically imitating examples. We present two-stage approach, namely retrieve-then-adapt. retrieval stage, index visual elements. For given information, transform it concrete query sampling from learned distribution about elements, then find appropriate our example library based similarity between indexes query. retrieved example, initial drafts replacing its content many cases, information cannot be perfectly fitted Therefore, further introduce adaption stage. Specifically, MCMC-like approach leverage recursive neural networks help adjust draft improve appearance iteratively, until satisfactory result obtained. implement widely-used proportion-related infographics, demonstrate effectiveness sample results expert reviews.