Published: 16 May 2023| Version 1 | DOI: 10.17632/vy4n28334m.1
Shravan Bhat,


The eye gaze images were captured using a smartphone device. Participants were asked to look at a camera and move their eye to the left, right, down, up and straight, resulting in a dataset of eye gaze images for analysis. The dataset consists of a total of 7,500 eye gaze images captured from 12 participants. Each image is labeled with one of five categories: left, right, up, down, or straight. The dataset is evenly distributed across these five categories, with 1,500 images for each category. The dataset of eye gaze images could be used to develop applications that can accurately predict the direction of a user's eye gaze based on the captured images. This could have applications in fields such as virtual reality, where accurate tracking of eye gaze is important for creating realistic experiences. Additionally, the findings of this study could be used to inform the design of assistive technology for individuals with disabilities that affect their ability to control their eye movements.


Steps to reproduce

The data was gather using a multiple recordings of diverse group of people, moving their eyes to the left, right, up, straight and down. Image frames of these videos were taken and labelled according to the direction of the eye. Each stimulus was presented for a fixed duration of 2 seconds, and participants were asked to maintain their gaze, with different pupil position for every direction of eye gaze. To ensure the quality of the data, we performed several quality control checks, such as excluding trials with blinks or excessive head movements, and visually inspecting the data for artifacts. The resulting dataset contains eye gaze images for each stimulus presentation, along with corresponding eye movement measures. To reproduce our research, one would just need a camera, a group of people and follow the same experimental protocol.


National Institute of Technology Karnataka