UniHair

This repository contains the official implementation for Towards Unified 3D Hair Reconstruction from Single-View Portraits.

Project Page | Arxiv

fastforward_x264.mp4

News

• TODO: release the rendering data.
• 2024.11.24: release the code!

Install

git clone --recursive https://github.com/PAULYZHENG/UniHair.git
cd UniHair

# change CUDA to 11.3 before create the env
conda env create -f environment.yml
conda activate unihair

sh setup.sh

Tested on Ubuntu 22 with torch 1.12.1 & CUDA 11.3 on 3090/4090.

Usage

Put images you want to test in data/img/

sh scripts/run_hairalign.sh
sh scripts/run_unihair.sh

Find your results in data/logs/

Acknowledgement

This repository is based on some excellent works, many thanks to all the authors!

• DreamGaussian
• gaussian-splatting and diff-gaussian-rasterization
• zero123 and zero123-hf
• 3DFFA_V2
• SAM

Citation

@article{zheng2024towards,
  title={Towards Unified 3D Hair Reconstruction from Single-View Portraits},
  author={Zheng, Yujian and Qiu, Yuda and Jin, Leyang and Ma, Chongyang and Huang, Haibin and Zhang, Di and Wan, Pengfei and Han, Xiaoguang},
  journal={arXiv preprint arXiv:2409.16863},
  year={2024}
}