Combining 3D Morphable Models: A Large scale Face-and-Head Model

This repository provides a MATLAB implementation of the CVPR 2019 Paper - Combining 3D Morphable Models: A Large scale Face-and-Head Model. It is implemented only till the Regression Matrix Calculation part.

Paper

Combining 3D Morphable Models: A Large scale Face-and-Head Model

Dependencies

• MATLAB
• Python >= 3.5
  • Numpy -> pip install numpy
  • Scipy -> pip install scipy

Usage

1. Cloning the repository

git clone https://github.com/nabeel3133/combining3Dmorphablemodels.git

2. Downloading the models

• LYHM: Liverpool York Head Model
  • After you have acquired LYHM, extract the lyhmPublic.zip and go to lyhmPublic/lyhmModels/, copy LYHM_male.mat and put it in the Regression Matrix Calculation folder.
• BFM09: Basel Face Model 2009
  • After you have acquired BFM, extract the BaselFaceModel.tgz and go to "PublicMM1" folder, copy 01_MorphableModel.mat and put it in the Regression Matrix Calculation folder.

3. Running the code

In order to run the code, launch MATLAB and follow the steps given below:

1. Open the file Steps1to3.m located in combining3Dmorphablemodels/Regression Matrix Calculcation and run it. (Keep in mind that there is a variable named total_heads on line #7, you can change it to as much head shape parameters you want the regression matrix to learn from).
2. After it is done executing, open the file nricp_run.m located in combining3Dmorphablemodels/Non Rigid Iterative Closest Point (NICP)/nricp-master/demos and run it. It will save Regression_Matrix.mat in the directory combining3Dmorphablemodels/Regression Matrix Calculcation.

4. Getting a Predicted Head

• In order to get a predicted head of a BFM face, you need to have the .obj file of the BFM face for which you want to predict the head. Make sure to have your .obj file named as Input_Face.obj. Locate to the directory combining3Dmorphablemodels/Prediction and run the following command:

  python head_prediction.py

• In order to generate a random BFM face from BFM model and then predict the head for that randomly generated face, locate to the directory combining3Dmorphablemodels/Prediction and run the following command:

  python head_prediction_rand_bfm.py

Both of the codes will output a file named Output_Head.obj located in the same directory. python head_prediction_rand_bfm.py will also save a file named Input_Face.obj which will contain the randomly generated BFM face.

Citation

If this work is useful for your research or if you use this implementation in your academic projects, please cite the following papers:

• Combining 3D Morphable Models: A Large Scale Face-And-Head Model

  @InProceedings{ploumpis2019combining,
  author = {Stylianos Ploumpis and Haoyang Wang and Nick Pears and William A. P. Smith and Stefanos Zafeiriou},
  title = {Combining 3D Morphable Models: A Large Scale Face-And-Head Model},
  booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  month = {June},
  year = {2019}
  }

• Optimal Step Nonrigid ICP Algorithms for Surface Registration:

  @InProceedings{amberg2007optimal,
  title={Optimal Step Nonrigid ICP Algorithms for Surface Registration},
  author={Amberg, Brian and Romdhani, Sami and Vetter, Thomas},
  booktitle={Computer Vision and Pattern Recognition, 2007. CVPR'07. IEEE Conference on},
  pages={1--8},
  year={2007},
  organization={IEEE}
  }