Abstract
Although variety of models of visual attention shift have been proposed, the ability to predict the shift of human attention, which is usually estimated from eye movements is limited. To improve models, it is necessary to include the effect of top-down processes. There is a report of an attention model that uses scene contexts evaluated based on global feature, a top-down cue to estimate the locations of gazes when searching targets such as persons. In this study, we added a process related to 3D visual perception to improve the model with contextual cueing.
