PyeLink and SyeLink are two complementary open-source Python tools for running eye-tracking experiments and parsing data from SR Research Ltd. EyeLink hardware. PyeLink simplifies experiment creation with plug-and-play support for multiple display backends (Pygame, PsychoPy, Pyglet) and enables two features not available in existing tools: recording data during calibration and validation phases, and simultaneously recording raw pupil/corneal-reflection data alongside calibrated gaze data. SyeLink processes the resulting enriched ASC files into structured JSON and CSV formats through both a Python API and a command-line interface, with built-in visualization of calibration and validation results.

Tools: PyeLink · SyeLink · Demo

We tested three lenses (16mm, 25mm, 35mm) on the EyeLink 1000 Plus to find the minimum eye-to-screen distance that still allows successful calibration while maintaining at least 15° of visual angle in the smallest dimension. The 16mm and 25mm lenses allowed the closest distances (330–335 mm to the screen top), while the 35mm lens required 457 mm. All configurations achieved average validation accuracy under 0.30°. Code and data: github.com/mh-salari/eyelink_min_distance_etra_lbw.

PyEtSimul is an open-source Python framework for simulating video-based eye trackers by generating synthetic eye features through geometric modeling. The framework allows flexible 3D positioning of eyes, cameras, and light sources, with controlled variation of eye anatomical features and camera properties. It offers both conic and spherical cornea models to choose from, supports non-circular pupil shapes, size-dependent pupil decentration, eyelid occlusion, and camera lens distortion, and enables comparison of gaze estimation algorithms across calibrated and uncalibrated settings.

Code: github.com/mh-salari/pyetsimul

In this study we examined how pupil size variations affect data quality in four head-mounted eye trackers — the Pupil Core, Pupil Neon, SMI ETG 2w, and Tobii Pro Glasses 2 — alongside the SR Research EyeLink 1000 Plus desktop tracker. By varying screen brightness to induce controlled pupil size changes, we measured the resulting apparent gaze shifts, accuracy, and precision. All head-mounted trackers exhibited the pupil-size artifact (PSA), with apparent gaze shifts ranging from 0.94° for the Pupil Neon to 3.46° for the Pupil Core. Precision effects were device-specific, with some trackers performing better in bright conditions and others in the dark.

In this study we introduced a standardized protocol for evaluating how lighting conditions affect head-mounted eye tracker accuracy. Testing SMI Eye Tracking Glasses 2 Wireless with 9 participants under low, moderate, and high illumination levels, we found that tracking accuracy is best when data recording occurs under the same lighting conditions used during calibration. Notably, calibration under moderate illumination provided the most consistent performance across all lighting environments.

We created the first comprehensive Persian SMS spam database — 4,389 labeled messages collected from various sources — and benchmarked several machine learning algorithms on it. A two-layer perceptron achieved the highest accuracy at 93.55%, establishing this database as a reliable starting point for Persian-language spam detection research.

Available on Civilica: civilica.com/doc/1670746

We introduced a deep-learning pipeline for automatic localization of caravanserais in satellite imagery. A database of 203 Iranian caravanserais was created and used to train YOLOv5 via transfer learning. To reduce false positives, misdetected locations were added to the database as a new class and the model was retrained. The final [email protected] on the test imagery was 91.43%.

Available on Civilica: civilica.com/doc/1955368