文章预览
本期带来本周在arXiv公开的论文11篇,涉及3D重建、3D生成、3D编辑、4D重建与生成。 一、3D重建 1.[2024-arXiv] Towards Unified 3D Hair Reconstruction from Single-View Portraits 作者机构:Yujian Zheng, et al. CUHK Shenzhen, Kuaishou Technology. 论文地址: https://arxiv.org/pdf/2409.16863 项目地址: https://unihair24.github.io/ 贡献: We propose a novel strategy to reconstruct textured 3D hair from single-view images, covering diverse hairstyles within a single pipeline. Our approach handles both simple un-braided styles, which are covered by state-of-the-art methods, as well as more complex braided styles such as ponytails, French braids, and buns. We contribute a large-scale synthetic hair dataset SynMvHair, covering various hairstyles, of calibrated multi-view images related to 2,396 collected 3D hair and 82,682 texture maps in total, which facilitates the learning of hair diffusion priors. It will be released to benefit the hair research.
………………………………