HairDeck generates animatable stylized hair cards from one image
Original title: HairDeck: Animation-Ready Stylized Hair Cards from a Single Image
On CHARM's test set it lowers Chamfer distance by 44% and card-orientation error by more than half, while running 33 times faster.
HairDeck treats a stylized hairstyle as an unordered set of hair cards and denoises every card together with a set diffusion transformer, using one token per card. From a single image it also generates the spring-bone chains that animate the cards and exports ready-to-use assets. The authors report it outperforms CHARM, image-to-3D pipelines, and strand-based methods.
- CHARM previously decoded stylized hair cards as a long autoregressive sequence of control points.
- Chamfer distance is 44% lower than CHARM on CHARM's own test set.
- Card-orientation error is more than halved, and inference is 33 times faster.
- Each generated card is driven by a spring-bone chain.
From the original article
1 min read at huggingface.co
The hair of stylized characters in games and avatars is built from hair cards: artist-placed ribbons whose geometry is the visible hair, animated by spring bones. Existing image-based methods produce realistic hair or a single fused mesh, while CHARM, the only generator of stylized hair cards, decodes a hairstyle autoregressively as a long sequence of control points.
By Yuze He, Yixiang Wei, Wang Zhao · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.