High-Fidelity Eye Animatable Neural Radiance Fields for Human Face

TL;DR

This paper introduces DeNeRF, a novel neural radiance field model that captures eye movements and eyeball rotations in face rendering, improving realism and gaze estimation accuracy from multi-view images.

Contribution

The paper proposes a new eye-aware face NeRF model that incorporates eyeball rotation and non-rigid periocular deformation using a canonical space transformation and an eye deformation field.

Findings

DeNeRF generates high-fidelity face images with accurate eye movements.

The model improves gaze estimation performance using rendered images.

Effective modeling of eyeball rotation enhances face rendering realism.

Abstract

Face rendering using neural radiance fields (NeRF) is a rapidly developing research area in computer vision. While recent methods primarily focus on controlling facial attributes such as identity and expression, they often overlook the crucial aspect of modeling eyeball rotation, which holds importance for various downstream tasks. In this paper, we aim to learn a face NeRF model that is sensitive to eye movements from multi-view images. We address two key challenges in eye-aware face NeRF learning: how to effectively capture eyeball rotation for training and how to construct a manifold for representing eyeball rotation. To accomplish this, we first fit FLAME, a well-established parametric face model, to the multi-view images considering multi-view consistency. Subsequently, we introduce a new Dynamic Eye-aware NeRF (DeNeRF). DeNeRF transforms 3D points from different views into a canonical space to learn a unified face NeRF model. We design an eye deformation field for the transformation, including rigid transformation, e.g., eyeball rotation, and non-rigid transformation. Through experiments conducted on the ETH-XGaze dataset, we demonstrate that our model is capable of generating high-fidelity images with accurate eyeball rotation and non-rigid periocular deformation, even under novel viewing angles. Furthermore, we show that utilizing the rendered images can effectively enhance gaze estimation performance.