Highlight Photo
Predicting Human Gaze
Saliency maps during a search for cups
The human fixations and saliency maps produced during a search for cups. Contextual guidance outperformed appearance modeling on this image, although it does not capture the two distinct regions where cups seem most likely two appear. Note than both human subjects and SUN’s appearance system are drawn to the cup-like objects above the fireplace. Yellow, green, and blue correspond to the top 10, 20, and 30 percent most salient regions respectively. This figure was selected by identifying images where the Contextual Guidance model and SUN most differed in their predictions.
Credits: Gary’s Unbelievable Research Unit (GURU), UCSD
