Yuze He, Yixiang Wei, Wang Zhao, Chuhao Chen, Yanning Zhou, Yong-Jin Liu (Tsinghua University, Carnegie Mellon University, The Chinese University of Hong Kong) · Hair card generation

HairDeck

HairDeck is a research method from Tsinghua University, Carnegie Mellon University and The Chinese University of Hong Kong. It turns a single image into stylised hair cards, together with the spring-bone chains that animate them. Each card is treated as one token, and a set diffusion transformer generates all the cards together. On the 100 test hairstyles used by CHARM, it reports a 44% lower Chamfer distance (0.0117 to 0.0066) and a card-orientation error of 16.8 degrees against 40.0 degrees for CHARM. Card generation is 33x faster, falling from 33.7 s to 1.01 s. The paper is on arXiv, the code is on GitHub, and models and a demo Space are on Hugging Face. No licence is stated.