Pytorch version of 'How far are we from solving the 2D & 3D Face Alignment problem? (and a dataset of 230,000 3D facial landmarks)'

For official torch7 version please refer to face-alignment-training [https://github.com/1adrianb/face-alignment-training]

This is a reinplement of training code for 2D-FAN and 3D-FAN decribed in "How far" paper. Please visit author's webpage [https://www.adrianbulat.com] or arxiv [https://arxiv.org/abs/1703.07332] for technical details.

Thanks for bearpaw's excellent work on human pose estimation [https://github.com/bearpaw/pytorch-pose] . And in this project, I reused a branch of helper function from pytorch-pose.

Pretrained models are available soon.

Requirments

• Install the latest Pytorch [http://pytorch.org], version 1.0.0.

Packages

• [https://www.scipy.org/][scipy]
  
• [https://pytorch.org][torchvision]
  
• [https://pypi.python.org/pypi/progress] progress (optional) for better visualization.

Train

1. Clone the github repository and install all the dependencies mentiones above.

     git clone https://github.com/lippman1125/pytorch_FAN
   

2. Download the LS3D-W dataset from the authors webpage (https://www.adrianbulat.com/face-alignment).

3. Download the 300W-LP annotations converted to t7 format by paper author from (https://www.adrianbulat.com/downloads/FaceAlignment/landmarks.zip).

4. We merge LS3D-W dataset and 300W-LP dataset together to train our model.

   cd data/LS3D-W
   tree -d -L 1
   

     |-- 300VW-3D                
     |-- 300W-Testset-3D         
     |-- 300W_LP                 
     |-- AFLW2000-3D-Reannotated 
     `-- Menpo-3D
   

   Validation set is testset of 300W-LP

5. Start to train:

     ./exp/train.sh
   

Test

1. Run the demo.

     python demo_video.py gpu
   

What's different?

• Train FAN only includes 2 Hourglass
• Total Params: 11.55M

Performance

Citation

  {
   @inproceedings{bulat2017far,
     title={How far are we from solving the 2D \& 3D Face Alignment problem? (and a dataset of 230,000 3D facial landmarks)},
     author={Bulat, Adrian and Tzimiropoulos, Georgios},
     booktitle={International Conference on Computer Vision},
     year={2017}
   }

Refenerce

• https://github.com/hzh8311/pyhowfar