AmbientEye: A Dataset for Pupil Segmentation under Natural Ambient Infrared Illumination
Researchers have introduced AmbientEye, a new dataset designed to advance pupil segmentation for eye-tracking in smart glasses. This dataset focuses on using passive infrared cameras under natural sunlight, avoiding the power consumption issues of active IR illumination. AmbientEye comprises over 2.6 million eye images from 35 participants, highlighting a significant performance drop for current segmentation algorithms in unconstrained outdoor settings. AI
IMPACT Enables development of more practical eye-tracking for AR/VR without active IR, potentially improving user interfaces and attention analysis.