Eye drawing with gaze estimation model

Abstract

Many research works in gaze-controlled interfaces have focused on target selection with eye gaze. For example, eye typing is built based on gaze patterns and language models. However only parts of these methods in target selection can be transferred to \nother domains. Unlike target selection in eye typing, eyedrawing target-selection is very small and the target could be less than 100 pixels target surface size. \nOverall, two kinds of eye drawing technique have already been proposed; a technique that applies free eye-paint [Tchalenko 2004; Gips and Olivieri 1996] and one that applies subtaskswitching \nbetween drawing and looking [Hornof et al. 2004]. \nThe former is successful at inspiring creativity and selfexpression \nbut cannot draw recognizable objects; and the latter \nis able to draw recognizable objects but cannot draw new objects \nin specific location on the screen.