Abstract
Little is known about gaze behavior in 360° naturalistic scenes. Much of what we know about gaze behavior comes from eye-tracking studies on fixed displays (computer monitors), where the fields of view are typically limited to ~20° visual angle, stimuli do not match real-world size, and images are not explored from a egocentric perspective via head-turns. Mobile eye-tracking studies overcome many of these limitations, affording a natural FOV, but the range of eye-tracking is typically limited to ~60°, and stimuli cannot be varied and controlled across participants. Here, we studied gaze-behavior in a headmounted display (FOV ~100°/120°), while subjects explored a set of naturalistic, 360° panoramic scenes from a first person perspective via head-turns. 20 individuals took part in this study. On each trial (40 trials; 20s each), participants viewed complex, real-world 360° panoramic scenes using a head-mounted display (resolution: 960×1080; field-of-view: ~100°; 75Hz). Viewing behavior was measured using an in-headset eye-tracker (120Hz; 5.7ms latency; 0.5° accuracy). During each trial (20s), participants actively explored a novel 360° panoramic scene, using head turns to change their viewpoint as they would in a real-world environment. We found that average saccade length during 360° scene exploration was much larger than expected: 24.3° visual angle +/= 2.49 STD, ~5° greater than the maximum possible measurements afforded by eye-tracking studies using a typical computer screen. Surprisingly, fixations were rarely at head center: only 4.6% (+/− 2.05 STD) fell within a 5° radius of head center on average. Lastly, we observed an “equator bias” in gaze behavior: fixations, raw gaze, and head direction were often directed participant within 20° of the equator of the panoramic scene 49% +/− 10.96 STD of the trial.