MVF-Net: Multi-View 3D Face Morphable Model Regression

Testing code for the paper.

MVF-Net: Multi-View 3D Face Morphable Model Regression.
Fanzi Wu*, Linchao Bao*, Yajing Chen, Yonggen Ling, Yibing Song, Songnan Li, King Ngi Ngan, Wei Liu. CVPR 2019.

Todo

• Code for rendering images.
• Replace image cropping scheme.

Installation

1. Python 2.7 (Numpy, PIL, scipy)

2. Pytorch 0.4.0, torchvision

3. face-alignment package from https://github.com/1adrianb/face-alignment. This code is used for face cropping and will be replaced by face detection algorithm in the future.

4. Model_shape.mat and Model_Expression.mat from 3DDFA.

Test

You can download the CNN model from here and copy it into data folder. Then you can test the model by:

python test_img.py --image_path ./data/imgs --save_dir ./result

If you are testing the code with your own images, please organize multiview images as:

folder
+--front.jpg
+--left.jpg
+--right.jpg

and change line 15 in test_img.py as:

crop_opt = True

Citation

If you find this work useful in your research, please cite:

@inproceedings{wu2019mvf,
  title={MVF-Net: Multi-View 3D Face Morphable Model Regression},
  author={Wu, Fanzi and Bao, Linchao and Chen, Yajing and Ling, Yonggen and Song, Yibing and Li, Songnan and Ngan, King Ngi and Liu, Wei},
  booktitle={CVPR},
  year={2019}
}