Gaze Estimation by Integrating Eye and Head Representation
Junhao Yue, Ping Lan, Yanxia Zhou, Zhicheng Dong · 2024
In this paper, we propose a dual-branch network model that requires both facial images and eye-label images as inputs, enhancing gaze estimation accuracy by capturing the intricate relationships among the head, face, and eyes. The strength of our network lies in its high accuracy of gaze estimation across different demographic groups. Our evaluation results indicate that our model performs exceptionally well on benchmark datasets Gaze360 and MPIIGaze, demonstrating strong competitiveness. We also conducted an ablation study to validate our approach.