Plausible 3D Face Wrinkle Generation Using Variational Autoencoders.

Area of Science:

• Computer Graphics
• Computer Vision
• Animation

Background:

• Realistic 3D facial modeling and animation are crucial for graphics, animation, and virtual reality.
• Generating fine-scale geometric wrinkles on animated 3D faces remains a significant challenge.

Purpose of the Study:

• To propose an end-to-end system for automatically augmenting coarse-scale 3D faces with synthesized fine-scale geometric wrinkles.
• To develop a method for generating plausible face wrinkles via a compact generative model.
• To introduce a pipeline for transferring synthesized wrinkles between faces with varying shapes and topologies.

Main Methods:

• Formulating wrinkle generation as a supervised task.
• Implicitly modeling the continuous space of face wrinkles using a compact generative model.
• Developing a pipeline for wrinkle transfer across different 3D face geometries and topologies.

Main Results:

• The system robustly synthesizes plausible fine-scale wrinkles on diverse coarse-scale 3D faces.
• The method effectively generates wrinkles for various facial expressions.
• Demonstrated successful transfer of synthesized wrinkles between faces of different shapes.

Conclusions:

• The proposed end-to-end system effectively addresses the challenge of generating realistic fine-scale wrinkles for 3D facial animation.
• The generative model and transfer pipeline enable plausible wrinkle synthesis and adaptation.
• This work advances the realism of 3D facial modeling and animation.