Homogenized Inflatables

This is Part I of the codebase for our SIGGRAPH paper, Computational Homogenization for Inverse Design of Surface-based Inflatables. The code is written primarily in C++, but it is meant to be used through the Python bindings.

Getting Started

C++ Code Dependencies

The C++ code relies on Boost, which must be installed separately.

The numerical solver depends on Catamari, which will be downloaded through the cmake file but require the meson build system. Alternatively one can choose to use the slower CHOLMOD/UMFPACK library, which must be installed separately.

The code also relies on several dependencies that are included as submodules: MeshFEM, libigl,

Finally, it includes a version of Keenan Crane's stripe patterns code modified to generate fusing curve patterns and fix a few issues with boundary handling.

macOS

You can install all the mandatory dependencies on macOS with MacPorts. When installing SuiteSparse, be sure to get a version linked against Accelerate.framework rather than OpenBLAS; on MacPorts this is achieved by requesting the accelerate variant, which is no longer the default. Simulations will run over 2x slower under OpenBLAS.

# Build/version control tools, C++ code dependencies
sudo port install cmake boost ninja meson cgal5
sudo port install SuiteSparse +accelerate
# Dependencies for jupyterlab/notebooks
sudo port install python39
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.1/install.sh | bash
nvm install 17 && nvm use 17
# Dependencies for `shapely` module
sudo port install geos

Ubuntu 20.04

A few more packages need to be installed on a fresh Ubuntu 20.04 install:

# Build/version control tools
sudo apt install git cmake ninja-build meson
# Dependencies for C++ code
sudo apt install libboost-filesystem-dev libboost-system-dev libboost-program-options-dev libsuitesparse-dev
# Dependencies (pybind11, jupyterlab/notebooks)
sudo apt install python3-pip npm
sudo npm install npm@latest -g
# Dependencies for `shapely` module
sudo apt install libgeos-dev

Obtaining and Building

Clone this repository recursively so that its submodules are also downloaded:

git clone --recursive https://github.com/jpanetta/Inflatables

Build the C++ code and its Python bindings using cmake and your favorite build system. For example, with ninja:

cd Inflatables
mkdir build && cd build
cmake .. -GNinja
ninja

Note that Catamari's performance is affected by the build settings, so to get best performance you'll want to choose the Release (not RelWithAssert) build type and enable MESHFEM_VECTORIZE.

Running the Jupyter Notebooks

The preferred way to interact with the inflatables code is in a Jupyter notebook, using the Python bindings. We recommend that you install the Python dependencies and JupyterLab itself in a virtual environment (e.g., with conda or venv).

pip3 install wheel # Needed if installing in a virtual environment
# Recent versions of jupyterlab and related packages cause problems:
#   JupyerLab 3.4 and later has a bug where the tab and status bar GUI
#                 remains visible after taking a viewer fullscreen
#   ipykernel > 5.5.5 clutters the notebook with stdout content
#   ipywidgets 8 and juptyerlab-widgets 3.0 break pythreejs
pip3 install jupyterlab==3.3.4 ipykernel==5.5.5 ipywidgets==7.7.2 jupyterlab-widgets==1.1.1
# If necessary, follow the instructions in the warnings to add the Python user
# bin directory (containing the 'jupyter' binary) to your PATH...

git clone https://github.com/jpanetta/pythreejs
cd pythreejs
pip3 install -e .
cd js
jupyter labextension install .

pip3 install matplotlib scipy networkx libigl setproctitle gmsh multiprocess
pip3 install shapely # dependency of the fabrication file generation

Launch JupyterLab from the root python directory:

cd python
jupyter lab

Now try opening and running an demo notebook, e.g., python/unit_cell_experiments/cosine_curve.ipynb.