Fig 1. Stimuli and experimental design.

A) Exemplars of the different object categories (cut-out objects from ImageNet validation set). 27 object categories were used in this experiment (subordinate level, based on ImageNet categories). In total, each category contained 10 exemplars. B) Stimuli were generated by placing the objects onto white, congruent and incongruent backgrounds (512*512 pixels, full-color). Backgrounds were sampled from the SUN2012 database [53]. For human participants, objects were downsized and placed in one of nine possible locations (3x3 grid). For DCNNs, objects were bigger and placed centrally. C) Participants performed on an object recognition task. At the beginning of each trial, a fixation-cross was presented in the center of the screen for 2000 ms, followed by an image. Images were presented in randomized sequence, for a duration of 32 ms, followed by a mask, presented for 300 ms. After the mask, participants had to indicate which object they saw, by clicking on one of 27 options on screen using the mouse. After 81 (⅓) and 162 (⅔) trials, there was a short break. D) Human performance (% correct) on the object recognition task. Participants performed best for segmented objects, followed by congruent and incongruent respectively. Error bars represent bootstrap 95% confidence intervals.