Source code for the paper:

Huang, Jingwei, Hao Su, and Leonidas Guibas. Robust Watertight Manifold Surface Generation Method for ShapeNet Models., arXiv preprint arXiv:1802.01698 (2018).

An advanced version has been released in this new repo.

We prepare the manifold data for 13 categories from ShapeNetCore. You can download them by running the following script.

wget http://download.cs.stanford.edu/orion/Shapenet_Manifold/categories.txt
wget -i categories.txt

For Linux and Mac users, run sh demo.sh to build and try the manifold example.

git clone --recursive -j8 git://github.com/hjwdzh/Manifold
cd Manifold
mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make

We take a triangle mesh "input.obj" and generate a manifold "output.obj". The resolution is the number of leaf nodes of octree. The face number increases linearly with the resolution.

./manifold input.obj output.obj [resolution (Default 20000)]

Our manifold software generates uniform manifold. For efficiency purpose, a mesh simplification can be used.

./simplify -i input.obj -o output.obj [-m] [-f face_num] [-c max_cost] [-r max_ratio]

Where:

  -m            Turn on manifold check, we don't output model if check fails
  -f face_num   Add termination condition when current_face_num <= face_num
  -c max_cost   Add termination condition when quadric error >= max_cost
  -r max_ratio  Add termination condition when current_face_num / origin_face_num <= max_ratio

./simplify -i input.obj -o output.obj -m -c 1e-2 -f 10000 -r 0.2

• Jingwei Huang

© Jingwei Huang, Stanford University

IMPORTANT: If you use this software please cite the following in any resulting publication:

@article{huang2018robust,
  title={Robust Watertight Manifold Surface Generation Method for ShapeNet Models},
  author={Huang, Jingwei and Su, Hao and Guibas, Leonidas},
  journal={arXiv preprint arXiv:1802.01698},
  year={2018}
}