Saliency for free: Saliency prediction as a side-effect of object recognition

Saliency is the perceptual capacity of our visual system to focus attention (i.e. gaze) on relevant objects instead background. So far, computational methods for saliency estimation required explicit generation a map, process which usually achieved via eyetracking experiments still images. This tedious that needs be repeated each new dataset. In current paper, we demonstrate possible automatically generate maps without ground-truth. approach, are learned as side effect object recognition. Extensive carried out both real and synthetic datasets demonstrated approach able accurate maps, achieving competitive results when compared with supervised methods.