PCFGaze: Physics-Consistent Feature for Appearance-based Gaze Estimation
This work addresses cross-domain gaze estimation for computer vision applications, offering a novel method to reduce overfitting and enhance accuracy.
The paper tackled the problem of understanding how gaze features relate to the physics of gaze by analyzing the gaze feature manifold, and the result was the PCFGaze framework, which improved cross-domain gaze estimation accuracy without extra training data.
Although recent deep learning based gaze estimation approaches have achieved much improvement, we still know little about how gaze features are connected to the physics of gaze. In this paper, we try to answer this question by analyzing the gaze feature manifold. Our analysis revealed the insight that the geodesic distance between gaze features is consistent with the gaze differences between samples. According to this finding, we construct the Physics- Consistent Feature (PCF) in an analytical way, which connects gaze feature to the physical definition of gaze. We further propose the PCFGaze framework that directly optimizes gaze feature space by the guidance of PCF. Experimental results demonstrate that the proposed framework alleviates the overfitting problem and significantly improves cross-domain gaze estimation accuracy without extra training data. The insight of gaze feature has the potential to benefit other regression tasks with physical meanings.