Geometry-Aware Appearance Learning for Generalized Gaze Estimation

Appearance-based gaze estimation has achieved remarkable progress in recent years. However, the inner geometry constraints of the eye and facial areas are not fully explored in existing methods, which limits the generalization and robustness of the model. In this letter, we propose a novel end-to-end framework for cross-domain gaze estimation by integrating the latent geometric representation into an appearance-based architecture. Specifically, we first exploit the 3DMM to fit unconstrained faces and eyes, which would generate adaptive normal information with the explicit 3D geometry prior. Then we joint the normal map and the corresponding RGB appearance information to infer the 3D gaze direction with carefully-designed spatial-frequency attention and local-global feature interaction modules. The key to our method is to integrate explicit 3D geometry representation into a 2D learning framework, which leads to a better trade-off between performance and efficiency. Experiments on both MPIIGaze and EyeDiap datasets demonstrate that the proposed method achieves the state-of-the-art accuracy of 3.56°and 5.10°separately, and also presents competing generalization on cross-domain evaluations.