Public code release for "Swept Volumes via Spacetime Numerical Continuation", presented at SIGGRAPH 2021 and authored by Silvia Sellán, Noam Aigerman and Alec Jacobson. Please note that while this implementation of our method is hereby released under MIT License, the method itself is pending a US patent filed in 2021 by Adobe Inc.
An up-to-date, maintained version of this code is implemented into our new python geometry processing library Gpytoolbox. See the swept volume documentation and examples here.
To install this library, please start by cloning the repository recursively
git clone --recursive git@github.com:sgsellan/swept-volumes.git
After this, compile in release mode using cmake following the usual steps:
cd swept-volumes
mkdir build
cd build
cmake -DCMAKE_BUILD_TYPE=Release ..
make all -j8
If all goes right, you'll find the executable swept-volumes-gui in the main repository directory. If not, we provide a MacOS precompiled binary in precompiled/osx/bin/swept-volumes-gui that you can move to the main repository directory.
You can begin by running ./swept-volumes-gui with no arguments. A viewer should appear showing a bunny:
You can then use W, E and R to flip between translation, rotation and scaling to move the bunny around.
Press "B" to bake a keyframe in. A copy of the bunny will appear fixed in the current position, and you can move the original bunny around to select your second keyframe:
Keep doing this until you are happy with the trajectory you've built:
Remember to press "B" to bake the last frame. Then, press "S" to run our algorithm to compute the swept volume. Wait for a couple of seconds (up to a minute in this example) and you'll see our output (this took 25s in my MacBook):
Press "V" to compare to a stamping algorithm with 10 stamps. Keep pressing "V" to see 100 stamps and 1000 stamps.
If you want to go back to our output, press "C". You can now close the viewer window, all the files have been saved to data/experiments/dd-mm-hh-mm in .obj format.
Once you have tried this demo, there are many arguments you can pass to our gui, in whichever order. -i path/to/shape.obj will change the input shape (any libigl-readable format is accepted and there are no requirements on manifold-ness or close-ness, default is data/bunny.obj), -e number changes the edge-length of the grid used in our algorithm (lower will be slower and more precise, default is 0.02). A sample call may be
./swept-volumes-gui -i path/to/your/shape.obj -e 0.01
A particularly interesting argument is -load, by which we can load the parameters and trajectory of a previous experiment. When using -load, the order of arguments does matter: every argument after -load will overwrite the loaded one. For example,
./swept-volumes-gui -i data/armadillo.obj -load data/experiments/experiment-name/
will ignore the -i data/armadillo.obj part and load whichever input was used in experiment-name experiment. However,
./swept-volumes-gui -load data/experiments/experiment-name/ -i data/armadillo.obj
will recreate experiment-name but with the armadillo as input. We can even add new keyframes when loading a previous experiment, or run the same experiment with a finer grid size:
./swept-volumes-gui -load data/experiments/experiment-name/ -e 0.01
This released code uses marching cubes as a surface generation algorithm, while our paper examples use dual contouring. I plan on adding dual contouring functionality to this repository soon when I manage to clean our original research code :-). All other elements in the algorithm are reproduced here exactly as were used to produce the examples in our paper.
Please do not hesitate to contact sgsellan@cs.toronto.edu if you find any issues or bugs in this code, or you struggle to run it in any way.