Researchers have developed a real-time eye-tracking system using a standard RGB webcam that estimates screen coordinates. The system integrates ETH-XGaze for gaze estimation with a nine-point calibration and polynomial ridge regression. It also incorporates a conditional residual correction based on head and pose features extracted by RTMPose to mitigate errors from head position changes. Preliminary tests with one participant showed a significant reduction in mean screen-coordinate error compared to existing methods, demonstrating the potential of combining user-specific calibration with pose-based correction for webcam gaze tracking. AI
IMPACT This research could lead to more accessible and accurate eye-tracking solutions for various applications, from accessibility tools to user interface control.
RANK_REASON This is a research paper detailing a novel system for eye-tracking. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →