The OpenSFEDS dataset (Open Sensor Fusion Eyes Data Set) is a novel dataset of 2.25M synthetic, time-synchronized eye-image pairs captured using a simulated RGB camera (640 x 480 px) and a set of 16 simulated photosensors placed on an on-axis sensor grid. The dataset has been created using a simulation framework based on UnityEyes, allowing for the synthesis of wide subjects’ appearance, illumination, and camera position variability. The dataset includes two subsets:

• OpenSFEDS-Static: 450K independent image pairs, with uniform gaze distribution covering ±25◦ of visual angle.
• OpenSFEDS-Temporal: 1.8K, 2s sequences of image pairs at 500 Hz with gaze extracted from real eye movement traces of the GazeBase dataset, featuring smooth pursuits, fixations, and saccades.

The paper introducing and describing the dataset is available here.

Detailed dataset description can be found here.

Data can be downloaded upon request. Please request access by sending an email to [email protected] or [email protected].

The dataset is licensed under the GNU General Public License v3.0 license.