A rule of thumb of icons' visual distinctiveness

EvIcon: Designing High‐Usability Icon with Human‐in‐the‐loop Exploration and IconCLIP

“…We expect that each icon I in $$\mathcal {I}$$ is associated with n text tags ($$\mathcal {T}_{I}={t_0, t_1,\ldots,t_{n-1}}$$) that represent the semantic and visual concepts of the icon such as 'search', 'next', 'television', and 'map'. We characterize the usability of an icon using common perceptual usability metrics including semantic distance , familiarity , and visual distinuguishability , which are commonly used by professional icon designers [MCdB01, Kur00, SM14]. However, these metrics of an icon are usually hard to evaluate mathematically from the icon image since the assessments of these metrics require extensive user testing to collect users' self‐reports and feedback.…”

“…Hence, due to the importance of these two indications for icon's usability [SABAG*05, MI09, SM14], we focus on providing icon designers with semantic distance and familiarity predictions on icon designs in this paper. Moreover, as an icon is usually designed and displayed within an icon set [Kur00], we also use visual distinguishability (i.e., ability to be easily recognized and differentiated from one another at a glance) as a critical usability criterion for designing an icon set [Kur00, LRFN04, SABAG*05, MS95]. The goal is to prevent users from confusing icons of different tags.…”

Presentation discovery

“…The first area involves descriptions of novel applications of icons spanning a spectrum from operating systems to complex multi-modal interfaces used to visualize scientific data [13,14]. The second area involves the evaluation of existing icons [4,10]. Task specific evaluations have included domains as diverse as medical data presentation [3], statistical interpretation and document indexing [2].…”