Awesome

WM3DR

Weakly-Supervised Multi-Face 3D Reconstruction

gif gif

Installation

To run the demo, the following requirements are needed.

pytorch >= 1.4
pytorch3d

Model

3DMM model used in this repo is from Deep3dPytorch, you should generate mSEmTFK68etc.chj file and put it into './BFM/'

Download the trained model final.pth in './model/'

Demo

python demo_video.py

Citation

If you find this project useful for your research, please use the following BibTeX entry.

@inproceedings{Zhang2021WeaklySupervisedM3,
   title={Weakly-Supervised Multi-Face 3D Reconstruction},
   author={Jialiang Zhang and Lixiang Lin and J. Zhu and S. Hoi},
   journal={arXiv: Computer Vision and Pattern Recognition},
   year={2021}
}