This repository contains the official PyTorch implementation of:

High-Fidelity Clothed Avatar Reconstruction from a Single Image

• Installation
• Dataset
• Training
• Inference

• CUDA=10.2
• Python = 3.7
• PyTorch = 1.6.0

1. Setup virtual environment:

conda create -n car python=3.7
conda activate car

# install pytorch
conda install -c pytorch pytorch=1.10.0 torchvision==0.7.0 cudatoolkit=10.2

# install pytorch3d 
pip install pytorch3d -f https://dl.fbaipublicfiles.com/pytorch3d/packaging/wheels/py37_cu102_pyt1100/download.html

# install trimesh  
conda install -c conda-forge rtree pyembree
pip install trimesh[all]

# install other dependencies
pip install -r requirement.txt

# install customized smpl code
cd smpl
python setup.py install
cd ../

If you use other python and cuda versions (default python3.7 cuda 10.2), please change the cuda version and python version in ./install.sh. If you use other pytorch version (default pytorch 1.6.0), please install pytorch3d according to the official install instruction official INSTALL.md.

2. Download smpl models from https://smpl.is.tue.mpg.de/, put them into models folder under ./data/smpl_related/models/smpl/

# CAR 
python -m apps.train -cfg configs/car-rp.yaml --gpu 0 

# ARCH* (*: re-implementation)
python -m apps.train -cfg configs/arch.yaml --gpu 0  

The results will be saved in ./out/.

• Download the pretrained models and put it in ./out/ckpt/ours-normal-1view/.
• Download extra data (PyMAF, ICON normal model, SMPL model) and put them to ./data.
• Run the following script to test example images in directory ./examples. Results will be saved in ./examples/results.

python -m apps.infer --gpu 0 -cfg configs/car-rp.yaml

@inproceedings{liao2023car,
  title     = {{High-Fidelity Clothed Avatar Reconstruction from a Single Image}},
  author    = {Liao, Tingting and Zhang, Xiaomei and Xiu, Yuliang and Yi, Hongwei and Liu, Xudong and Qi, Guo-Jun and Zhang, Yong and Wang, Xuan and Zhu, Xiangyu and Lei, Zhen},
  booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
  month     = {June},
  year      = {2023},
}