Analysis of exploratory patterns is a very debated issue in literature but, unfortunately, relevant works in the state of the art leverage data acquired by using a remote or wearable eye tracker. This page contains information and links about the dataset UserWGaze, which represents a dataset presenting facial data of 8 subjects instead of gaze track extracted by an eye tracker. The dataset was built by involving 8 different subjects (7 men, 1 woman): each subject was asked to position himself in front of a 28-inch monitor at an approximate distance of about 70 cm, and a webcam recorded his face while 5 short clips of about 20 seconds each were projected on the screen.
The subjects were acquired while watching each clip during three different sessions. Each session has been performed with at least 24 hours interval. In particular, the five clips used as stimuli were selected among those available at http://www.inb.uni-luebeck.de/tools-demos/gaze.
These are the name of the files in the third part repository: ‘beach’, ‘breite_strasse’, ‘bridge_1’, ‘bumblebee’ and ‘doves’.Associated gaze information, extracted by using an algorithmic pipeline that leverages state of art algorithms for face detection, facial landmark extraction, and gaze track extraction are included. Also, other soft biometrics traits regarding the age and gender of the participants have been inserted in the dataset, to provide the possibility to use the proposed dataset also for other forms of soft biometrics authentication. The dataset consists of video files of the recorded sessions and a command separated values (CSV) structure where each line contains information like age, gender, identity and temporal gaze tracks of the participant. The dataset has been introduced in the paper
“Investigating Ocular Biometrics Recognition by Modelling Human Attention During Video Observations” by Dario Cazzato, Pierluigi Carcgnì, Claudio Cimarelli, Holger Voos, Cosimo Distante and Marco Leo that is under review for publication in the Elsevier Journal on Image and Vision Computing.
The dataset is released under an Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) license (https://creativecommons.org/licenses/by-nc/4.0/).
For more information please contact
