EG-SNIK: A Free Viewing Egocentric Gaze Dataset and Its Applications

Egocentric vision data captures the first person perspective of a visual stimulus and helps study gaze behavior in more natural contexts. In this work, we propose new dataset collected free viewing style with an end-to-end processing pipeline. A group 25 participants provided their information wearing Tobii Pro Glasses 2 set up at museum. The stream is post-processed for handling missing or incoherent information. corresponding video clipped into 20 videos to museum exhibits compensated user’s unwanted head movements. Then, I-VT algorithm used identify eye movements as fixations saccades. Based on velocity directional shifts eye, classifies either Representative scanpaths are built by generalizing multiple viewers’ gazing styles all exhibits. application our demonstrated characterizing inherent dynamics using state trajectory estimator based ancestor sampling (STEAS) model solving classification retrieval problems.