Learning to Model Task-Oriented Attention

Figure 2

(a) The percentage of gaze points within the center 11% and 25% of the images, which is displayed by blue. Meanwhile, red shows the percentage of target gaze points. Obviously, in the dog category, the percentage of target gaze points is more than chair category. ((b) and (c)) Dog’s and chair’s average saliency map containing all the gaze points, which indicates a bias to the center of the image.

(a)

(b)

(c)