Gaussian Haircut creates strand-based hair geometry from a video.
Egor Zakharov et al.
Experiments with 3D Gaussian Splatting are evolving to produce increasingly more accurate models. Researchers from ETH Zurich, Max Planck Institute for Intelligent Systems, Meta, and Technical University of Darmstadt presented a new method of reconstructing hair using the technology.
3D Gaussian Splatting is a rendering technique that leverages 3D Gaussians to represent the scene, thus allowing one to synthesize 3D scenes out of 2D footage. Simply put, it takes samples of images and turns them into 3D scenes without creating meshes by converting a point cloud to Gaussians using machine learning. You can learn more about it here.
Gaussian Haircut uses a dual representation of hair strands and 3D Gaussians to produce realistic strand-based reconstructions from a video, which takes a few hours on a single GPU.
"In contrast to recent approaches that leverage unstructured Gaussians to model human avatars, our method reconstructs the hair using 3D polylines, or strands. This fundamental difference allows the use of the resulting hairstyles out-of-the-box in modern computer graphics engines for editing, rendering, and simulation."
The method relies on unstructured Gaussians to generate multi-view data to supervise the fitting of hair strands, and the hairstyle is represented in the form of the"strand-aligned 3D Gaussians."
There are several stages in the process: first, the method takes multi-view images and turns them into unstructured Gaussians. Then, it "fits" the strands to structure the strands and outputs a strand-based hairstyle.
Hairstyles created with Gaussian Haircut can be imported into physically rendered and simulated virtual environments.
Egor Zakharov et al.
If you want to know the technical details of the method, find the project page here.
Also, join our 80 Level Talent platform and our Telegram channel, follow us on Instagram, Twitter, LinkedIn, TikTok, and Reddit, where we share breakdowns, the latest news, awesome artworks, and more.