Point2Mesh is a technique for reconstructing a surface mesh from an input point cloud. This approach "learns" from a single object, by optimizing the weights of a CNN to deform some initial mesh to shrink-wrap the input point cloud. The argument for going this route is: since the (local) convolutional kernels are optimized globally across the entire shape, this encourages local-scale geometric self-similarity across the reconstructed shape surface.

The code was written by Rana Hanocka and Gal Metzer.

• Clone this repo:

git clone https://github.com/ranahanocka/point2mesh.git
cd point2mesh

• Relies on PyTorch version 1.4 (or 1.5) and PyTorch3D version 0.2.0.
  Install via conda environment conda env create -f environment.yml (creates an environment called point2mesh)

This code relies on the Robust Watertight Manifold Software. First cd into the location you wish to install the software. For example, we used cd ~/code. Then follow the installation instructions in the Watertight README. If you installed Manifold in a different path than ~/code/Manifold/build, please update options.py accordingly (see this line)

Download our example data

bash ./scripts/get_data.sh

First, if using conda env first activate env e.g. source activate point2mesh. All the scripts can be found in ./scripts/examples. Here are a few examples:

bash ./scripts/examples/giraffe.sh

bash ./scripts/examples/bull.sh

bash ./scripts/examples/tiki.sh

bash ./scripts/examples/noisy_guitar.sh

... and more.

To run all the examples in this repo:

bash ./scripts/run_all_examples.sh

You should provide an initial mesh file. If the shape has genus 0, you can use the convex hull script provided in ./scripts/process_data/convex_hull.py

If you find this code useful, please consider citing our paper

@article{Hanocka2020p2m,
  title = {Point2Mesh: A Self-Prior for Deformable Meshes},
  author = {Hanocka, Rana and Metzer, Gal and Giryes, Raja and Cohen-Or, Daniel},
  year = {2020},
  issue_date = {July 2020}, 
  publisher = {Association for Computing Machinery}, 
  volume = {39}, 
  number = {4}, 
  issn = {0730-0301},
  url = {https://doi.org/10.1145/3386569.3392415},
  doi = {10.1145/3386569.3392415},
  journal = {ACM Trans. Graph.}, 
}

If you have questions or issues running this code, please open an issue.