Extras

In addition to the video stimuli and the raw eye-movement traces, we have provided:

• Visualization clips, which consist of the original video clips onto which we superimposed the eye movement traces available for each clip;
• Thumbnails for each video clip (one representative image);
• Parsing of the eye-movement traces and addition of additional fields which indicate when the subject is in fixation, smooth pursuit, saccade, etc.
• Predictions, in the form of salience maps, from three computational models which we ran on these stimuli: saliency (Itti et al., 1998; 2000; 2001), Bayesian surprise (Itti & Baldi, 2005; 2006), and a simple control measuring local pixel variance (Itti & Baldi, 2006).
• Additional visualizations, where each clip is displayed alongside the dynamically computed salience maps (for each of the three model variants), with human eye movement traces superimposed;
• Examples of perl/shell scripts which we used to process the video clips, compute the saliency maps, and compare the saliency maps to the eye-movement traces;
• Examples of Matlab scripts to compute statistics and display results.