The Attention System VOCUS

In this section, we present the goal-directed visual attention system VOCUS (Visual Object detection with a CompUtational attention System) (cf. Fig. 2). With visual attention we mean a selective search-optimization mechanism that tunes the visual processing machinery to approach an optimal configuration [21]. VOCUS consists of a bottom-up part computing data-driven saliency and a top-down part enabling goal-directed search. The global saliency is determined from bottom-up and top-down cues. In the following, we first describe the computation of the bottom-up and then of the top-down saliency.

Figure 2: The goal-directed visual attention system VOCUS with a bottom-up part (left) and a top-down part (right). In learn mode, target weights are learned (blue line arrows). These are used in search mode (red short arrows).