Mobile gaze tracking often relies on appearance-based 2D gaze estimation using the device's front-facing RGB camera. This approach makes eye tracking more accessible by avoiding the need for dedicated hardware, but it may be sensitive to movement.
A new paper posted on arXiv quantifies the impact of motion on this method in real-world mobile interactions. The study focuses on how typical handling of a phone or tablet affects the accuracy of gaze estimates, rather than assuming a static setup.
The findings underscore motion as a key challenge for practical mobile gaze tracking. By measuring this degradation, the work provides a quantitative basis for understanding the limits of current appearance-based methods and for guiding future improvements.