3dlandmarkdetection

This code demonstrates the functionality as shown in:

B. Berends, F. Bielevelt, R. Schreurs, S. Vinayahalingam, T. Maal, and G. de Jong, “Fully automated landmarking and facial segmentation on 3D photographs,” arxiv, 2023. arXiv:2309.10472

Please cite this paper when you use this work (see License).

IMPORTANT: Additional License information

This work is published under the Apache License Version 2.0. There are additional conditions for using this model beyond non-commercial reserach or educational use. This model has partially been trained using the HeadSpace Dataset. If you would use this work for any other purpose other than non-commercial or educational use, you must request permission at the HeadSpace team (see website).

Method overview

Installation

For the installation we have the following template instructions for a python 3.10 instruction.

1. Install the python environment and requisites

conda create -n landmarking python=3.10
conda activate landmarking

2. (Optional) Some users reported a ClobberError regarding libnpp-dev when installing torch. You can fix this by: conda install -c nvidia libnpp-dev
3. The following packages need to be installed as following

conda install pytorch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 pytorch-cuda=11.7 -c pytorch -c nvidia
conda install -c anaconda pandas
conda install -c conda-forge scipy
conda install -c anaconda scikit-learn
conda install -c conda-forge tqdm
pip3 install open3d
pip3 install pymeshlab
pip3 install robust-laplacian
pip3 install potpourri3d

4. Clone the repository to a local folder (e.g. landmarking)
5. Clone the following repository https://github.com/Goblaski/diffusion-net-cuda link and put it in the local folder under diffusion-net
6. Navigate to diffusion-net and type python setup.py install
7. Clone the following repository https://gitlab.kuleuven.be/mirc/meshmonk link and put it in the local folder under meshmonk
8. Follow the installation instructions of meshmonk You should now be set to run the demo.

Folder structure

After installation you shoulw have the following folder structure:

/Landmarking/common_functions/
/Landmarking/diffusion-net/
/Landmarking/initial_landmarking/
/Landmarking/meshmonk/
/Landmarking/meshmonk_script/
/Landmarking/models/
/Landmarking/refined_landmarking/

Demo

1. Navigate to the root folder (e.g landmarking)
2. run python prepare_demo.py
3. run python Initial_landmarking.py Optionally you can change --input folder and add --export_realigned_mesh
4. In Matlab, open meshmonk_script/MeshMonkFolder.m and run.
5. run python Refined_landmarking.py Optionally you can change --input folder , add --export_realigned_mesh , and/or --save_segmented_mesh.