Comparison in performance between standard computational models and human observers towards the recognition pattern of people and sport categories.

Again, two scales are analyzed here: “L”- the largest level, “S”—the smallest level. SALIENCY model performed at a chance level that validated that low-level cues were almost not informative in the dataset. GIST model that represented global layout statistics was close to explaining human data when the target object (the human-object pair for sport category) is relatively small.