HairDeck generates stylized hair cards from one image, with spring-bone chains so the hair can be animated.
Core idea: Each hair card is a single token, and a set diffusion transformer denoises all cards together, instead of decoding a long sequence of control points the way CHARM does.
Output: From a single image it produces the cards and the spring-bone chains that drive them, exporting to ready-to-use downstream formats.
Claimed results: On CHARM's test set, it lowers Chamfer distance by 44%, cuts card-orientation error by more than half, and runs 33 times faster.
Comparisons: The paper says it beats CHARM, image-to-3D pipelines and strand-based methods on single images.
No discussion yet. The numbers are the paper's own claims, and the project page link is in the abstract if you want to check the demos.
Sign in to comment.
TL;DR of the paper, generated automatically.
HairDeck generates stylized hair cards from one image, with spring-bone chains so the hair can be animated.
No discussion yet. The numbers are the paper's own claims, and the project page link is in the abstract if you want to check the demos.