Establishing Robust Retinal Eye Tracking: A Weakly Supervised Algorithmic Framework

2026-05-09

AI Summary

Existing retinal eye-tracking methods predominantly rely on conventional template matching, which suffers from insufficient robustness under feature variations and realistic imaging conditions. This work proposes the first weakly supervised, learning-based retinal eye-tracking framework that integrates deep neural networks with image registration techniques, substantially reducing reliance on densely annotated data. Evaluated across six subjects, the method achieves a 95th-percentile eye movement error of less than 0.45 degrees, demonstrating high accuracy, strong robustness, and excellent generalization capability.

Abstract

Retinal image-based eye tracking is widely used in ophthalmic imaging and vision science, and is a promising path to deliver higher gaze accuracy than the pupil- and cornea-based approaches commonly used in modern AR/VR devices. Nevertheless, existing retinal tracking algorithms still primarily rely on classical template-matching registration, which can be insufficiently robust to retinal feature variability and real-world imaging conditions. In this work, we propose a novel weakly-supervised, learning-based framework for robust retinal eye tracking. Initial studies demonstrate high accuracy, achieving the 95th-percentile gaze error < 0.45 deg across a cohort of 6 participants.

Problem

• retinal eye tracking
• robustness
• feature variability
• imaging conditions
• gaze accuracy

Innovation

• weakly-supervised learning
• retinal eye tracking
• gaze accuracy
• robust registration
• ophthalmic imaging