Fast Tetrahedral Meshing in the Wild¶
💡💡💡 If you are interested in the algorithm details, please refer to our paper first. We provide plenty of examples and statistics in the paper.
💡💡💡 Check our license first.
💡💡💡 Our algorithm is robust both in theory and in practice. If you do find fTetWild crash (on your laptop), please test it (on cluster) with more resource given. The most complex model I tested requires >32GB memory.
💡💡💡 The orientation of input faces is as important as the position of their vertices. Our algorithm is faithful to the input face position and orientation check Tetwild
Here is pre-generated tetmeshes and the extracted surface meshes for research-purpose usage. Please kindly cite our paper when using our pre-generated data.
- Input: Thingi10k
Installation via CMake¶
Our code was originally developed on MacOS and has been tested on Linux and Windows. We provide the commands for installing fTetWild in MacOS:
- Clone the repository into your local machine:
git clone https://github.com/wildmeshing/fTetWild.git
- Compile the code using cmake (default in Release mode):
You may need to install
gmp before compiling the code. You can install them via
- homebrew on mac:
brew install gmp
- Package manager on Unix:
sudo apt-get install gmp
Conda on Windows:
conda install -c conda-forge mpfr
Check the installation:
The inputs of our software are triangle surface meshes in
.mesh/.msh format output. The default output format is
.msh with minimum dihedral angle recorded as element scalar field, which can be visualized by software Gmsh. You can use
pymesh/ for read and write
Our software is quite easy to use. Basically, users only need to provide a surface triangle mesh as input and our mesher would output a tetrahedral mesh by using default settings. If you want to customize your own tetmeshes, we also provide some options.
- Envelope of size epsilon
Using smaller envelope preserves features better but also takes longer time. The default value of epsilon is b/1000, where b is the length of the diagonal of the bounding box.
- Ideal edge length
Using smaller ideal edge length gives a denser mesh but also takes longer time. The default ideal edge length is b/20
- Filtering energy
Our mesher stops optimizing the mesh when maximum energy is smaller than filtering energy. Thus, larger filtering energy means less optimization and sooner stopping. If you do not care about quality, then give a larger filtering energy would let you get the result earlier. The energy we used here is conformal AMIPS whose range is from 3 to +inf. The default filtering energy is 10.
💡 We suggest not to set filtering energy smaller than 8 for complex input.
- Maximum number of optimization passes
Our mesher stops optimizing the mesh when the maximum number of passes is reached. The default number is 80.
- Smoothing open regions
Our method can fill gaps and holes but the tetmesh faces on those parts could be bumpy. We provide users an option to do Lapacian smoothing on those faces to get a smoother surface.
Command Line Switches¶
Our software supports usage via command line or via a C++ function wrapper. Here is an overview of all command line switches:
Usage: ./FloatTetwild_bin [OPTIONS]
-h,--help Print this help message and exit
-i,--input TEXT:FILE Input surface mesh INPUT in .off/.obj/.stl/.ply format. (string, required)
-o,--output TEXT Output tetmesh OUTPUT in .msh format. (string, optional, default: input_file+postfix+'.msh')
-l,--lr FLOAT ideal_edge_length = diag_of_bbox * L. (double, optional, default: 0.05)
-e,--epsr FLOAT epsilon = diag_of_bbox * EPS. (double, optional, default: 1e-3)
--log TEXT Log info to given file.
--level INT Log level (0 = most verbose, 6 = off).
-q,--is-quiet Mute console output. (optional)
--csg TEXT:FILE json file containg a csg tree
--disable-wn Disable winding number.
--max-threads UINT maximum number of threads used
We used several useful libraries in our implement, testing, and rendering listed as follows. We would like to especially thank their authors for their great work and publishing the code.