Analysis of scores, datasets, and models in visual saliency prediction Supplementary Material

We compare 32 saliency models. Details of our model comparison setup and compared models are provided in this supplement. We report eye movement prediction accuracy of models using three scores: Shuffled AUC (sAUC), Correlation Coefficient (CC), and Normalized Scanpath Saliency (NSS). Additionally, we show results of statistical significance (statistical tests) of models using t-test. Results are reported over four widely-used and benchmark eye-tracking datasets which are publicly available. In the following, we explain statistics of employed datasets.