StrandHead: Text to Strand-Disentangled 3D Head Avatars Using Hair Geometric Priors
Xiaokun Sun, Zeyu Cai, Zhenyu Zhang, Ying Tai, Jian Yang
2024-12-17
Summary
This paper talks about StrandHead, a new method for creating 3D head avatars that can generate realistic hairstyles based on text descriptions, allowing for detailed customization.
What's the problem?
Current methods for generating avatars often struggle to accurately represent hair, leading to unrealistic or overly simplified hairstyles. This is because they typically use general representations that don't capture the complexity of real hair, making it difficult to create avatars that reflect individual personalities and styles.
What's the solution?
StrandHead addresses this issue by introducing a novel approach that generates hair as distinct strands rather than a generalized shape. It uses a combination of 2D generative models to create realistic hair without needing 3D data for training. The method leverages specific features about hair shapes and styles, allowing it to produce high-quality, detailed hairstyles that match the text prompts provided by users.
Why it matters?
This research is important because it enhances the ability to create personalized and realistic 3D avatars, which can be used in various applications like video games, virtual reality, and animation. By improving how avatars represent hair, StrandHead can help users express their individuality more effectively in digital environments.
Abstract
While haircut indicates distinct personality, existing avatar generation methods fail to model practical hair due to the general or entangled representation. We propose StrandHead, a novel text to 3D head avatar generation method capable of generating disentangled 3D hair with strand representation. Without using 3D data for supervision, we demonstrate that realistic hair strands can be generated from prompts by distilling 2D generative diffusion models. To this end, we propose a series of reliable priors on shape initialization, geometric primitives, and statistical haircut features, leading to a stable optimization and text-aligned performance. Extensive experiments show that StrandHead achieves the state-of-the-art reality and diversity of generated 3D head and hair. The generated 3D hair can also be easily implemented in the Unreal Engine for physical simulation and other applications. The code will be available at https://xiaokunsun.github.io/StrandHead.github.io.