Self-supervised Deformation Modeling for Facial Expression Editing

TL;DR

This paper introduces a self-supervised, two-step neural network for facial expression editing that explicitly models facial motion through image deformation and texture generation, achieving more realistic results.

Contribution

It proposes a novel physically-based, self-supervised system that disentangles motion and texture editing, improving facial expression editing without needing deformation annotations.

Findings

Outperforms state-of-the-art in qualitative evaluations.

Achieves higher quantitative scores in expression editing.

Removes the need for ground truth deformation data.

Abstract

Recent advances in deep generative models have demonstrated impressive results in photo-realistic facial image synthesis and editing. Facial expressions are inherently the result of muscle movement. However, existing neural network-based approaches usually only rely on texture generation to edit expressions and largely neglect the motion information. In this work, we propose a novel end-to-end network that disentangles the task of facial editing into two steps: a " "motion-editing" step and a "texture-editing" step. In the "motion-editing" step, we explicitly model facial movement through image deformation, warping the image into the desired expression. In the "texture-editing" step, we generate necessary textures, such as teeth and shading effects, for a photo-realistic result. Our physically-based task-disentanglement system design allows each step to learn a focused task, removing the need of generating texture to hallucinate motion. Our system is trained in a self-supervised manner, requiring no ground truth deformation annotation. Using Action Units [8] as the representation for facial expression, our method improves the state-of-the-art facial expression editing performance in both qualitative and quantitative evaluations.