Portrait3D

This is the official code repository for our SIG'24 paper:

Portrait3D: Text-Guided High-Quality 3D Portrait Generation Using Pyramid Representation and GANs Prior

ACM Transactions on Graphics (Proceedings of SIGGRAPH 2024)

Yiqian Wu, Hao Xu, Xiangjun Tang, Xien Chen, Siyu Tang, Zhebin Zhang, Chen Li, Xiaogang Jin*

<div align="center">

</div>

https://github.com/oneThousand1000/Portrait3D/assets/32099648/bc5721db-6edb-4d6a-a88f-07305476cd67

Requirements

1. Tested on Python 3.8
2. At least 12 GB of memory
3. Tested on NVIDIA RTX 3080Ti with 12 GB of memory (Windows, 1.5h per portrait)
4. Tested on NVIDIA RTX 4090 with 24 GB of memory (Linux, 0.5h per portrait)
5. CUDA>=11.6

Installation

Clone this repo to $PROJECT_ROOT$.

Create environment

cd $PROJECT_ROOT$
conda env create -f environment.yaml
conda activate text_to_3dportrait

Torch and torchvision Installation

pip install torch==1.12.1+cu116 torchvision==0.13.1+cu116 -f https://download.pytorch.org/whl/torch_stable.html

OSMesa Dependencies (For Linux)

sudo apt install  libosmesa6  libosmesa6-dev

Installing Additional Requirements

pip install -r requirements.txt

kaolin Installation

pip install kaolin==0.13.0 -f https://nvidia-kaolin.s3.us-east-2.amazonaws.com/torch-1.12.1_cu116.html

Stable-diffusion Installation

cd stable-diffusion
pip install -e .
cd ..

SMPL Model Setup

1. Download SMPL_python_v.1.0.0.zip (version 1.0.0 for Python 2.7 (female/male. 10 sha