Toward 3D Face Reconstruction in Perspective Projection: Estimating 6DoF Face Pose From Monocular Image

This repository contains the official PyTorch implementation of:

Toward 3D Face Reconstruction in Perspective Projection: Estimating 6DoF Face Pose From Monocular Image

Installation

Setup python environment

CUDA>=10.1
Python = 3.7
PyTorch >= 1.7.1

Install other dependencies

pip install -r requirement.txt

Install libs

Please follow ./lib/mesh/README.md.
Please follow ./lib/mesh_p/README.md.
Please follow ./lib/Sim3DR/README.md.

Demo

Download our pretraind model from here and put it in the folder ./checkpoint/run1/latest_net_R.pth.

$ python demo.py

Download Dataset

** Download ARKitFace dataset**

Please contact email cbsropenproject@outlook.com. We will send you an agreement for signing, and after our verification, the download link for ARKitFace dataset will be sent to you. Please ensure compliance with agreement and do not use this dataset for any non research purposes.

Then put the download dataset in the folder ./dataset/ARKitFace.

Testing and training

Download our pretraind model from here and put it in the folder ./checkpoint/run1/latest_net_R.pth.

Testing

$ python -u test.py --csv_path_test test.csv

Training

$ python -u train.py  --csv_path_train train.csv --csv_path_test test.csv

Citation

If you find our work usefull in your research, please use the following BibTex entry.

@ARTICLE{10127617,
  author={Kao, Yueying and Pan, Bowen and Xu, Miao and Lyu, Jiangjing and Zhu, Xiangyu and Chang, Yuanzhang and Li, Xiaobo and Lei, Zhen},
  journal={IEEE Transactions on Image Processing}, 
  title={Toward 3D Face Reconstruction in Perspective Projection: Estimating 6DoF Face Pose From Monocular Image}, 
  year={2023},
  volume={32},
  number={},
  pages={3080-3091},
  doi={10.1109/TIP.2023.3275535}}