Self-Supervised Monocular 3D Face Reconstruction by Occlusion-Aware Multi-view Geometry Consistency(ECCV 2020, unfinish)

This is an official python implementation of MGCNet.

Result

1. video

2. image

3. full video Full video can be seen in [YouTube] https://www.youtube.com/watch?v=DXzkO3OwlYQ

Running code

1.Model

1. 3dmm + network weight https://drive.google.com/file/d/1RkTgcSGNs2VglHriDnyr6ZS5pbnZrUnV/view?usp=sharing Extrack this file to /MGCNet/model
2. pretain https://drive.google.com/file/d/1jVlf05_Bm_nbIQXZRfmz-dA03xGCawBw/view?usp=sharing Extrack this file to /MGCNet/pretain

2.Data

1. data https://drive.google.com/file/d/1Du3iRO0GNncZsbK4K5sboSeCUv0-SnRV/view?usp=sharing Extrack this file to /MGCNet/data
2. data: landmark ground truth
3. data: skin prob

3.Requirement

1. as the requirement.txt

4.Time

The run time is ~0.02s with GPU(GeForce GTX 2070) and ~0.8s with CPU.

5.Testing

6.Training

1. cmd

Useful tools

1. comparison rendering

Citation

If you use this code, please consider citing:

eccv

Contacts

Please contact jiaxiang.shang@gmail.com or open an issue for any questions or suggestions.

Acknowledgements