We propose a new meshing algorithm to generate a surface with correct topology for the output points of a point network. GAMesh can be used both in post-processing to mesh the output points or to train the point network to directly optimize the vertex positions of the final 3D mesh. Unlike traditional surface reconstruction techniques (like Ball-Pivot Algorithm & Screened Poisson Reconstruction), GAMesh is independent of both the density and distribution of points and gaurantees correct geometry and topology of the final surface. We show its applications in single-view shape prediction, fair evaluation of point networks and reconstructing surfaces for networks which output sparse point clouds.

coming soon

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

@inproceedings{agarwal2020GAMesh,
    title={GAMesh: Guided and Augmented Meshing for Deep Point Networks},
    author={Agarwal, Nitin and Gopi, M},
    booktitle={3DV},
    year={2020}}

Our code is released under MIT license (see License file for details)

Please contact Nitin Agarwal if you have questions or comments