Fig. 3
From: Finding any Waldo with zero-shot invariant and efficient visual search
Experiment 1 (Object arrays). a Example target and search images for one trial. The green circle showing the target position was not shown during the actual experiment. Due to copyright, the picture of a sheep in the target image is not the actual one used in this experiment. b Attentional map overlaid on the search image and sequence of fixations for the IVSN model. c Sequence of fixations for 5 subjects (out of 15 subjects). The number above each subplot shows the number of fixations when the target was found. d Distribution of reaction times for the first fixation (see Supplementary Figure 2 for the corresponding distributions for subsequent fixations). Gray lines show the reaction time distributions for 5 individual subjects, the black line shows the average distribution. The median reaction time was 248 ms (SD = 130 ms, vertical dashed line). e Cumulative performance as a function of fixation number for humans (red), IVSN oracle model (blue), and chance model (dashed line). Error bars denote SEM, n = 15 subjects (see Supplementary Figure 4 for comparisons against other models)
