Point2Tet

Point2Tet is a novel deep neural network to learn Tetrahedral Meshed from 3d point clouds.

Tetrahedral Mesh of a Cube

Getting Started

• Clone or download this repo.
• Install dependecies (requires PyTorch).
• Run main.py using Python (version 3.x.x).

Important CLI Parameters

1. --name: The name of the generated folder in which the data will be saved
2. --input_cube: Controls resolution. Path to the file containing the file to load as the initial tetrahedral mesh (a cube is assumed in .tet format)
3. --input_filled_pc: Path to the filled point cloud to use as input. (.obj format)
4. --input_surface_pc: [Optional] Path to the surface point cloud (without normals) to use as additional input. (.obj format)

Examples

1. sphere: python main.py --name sphere --init_cube ../objects/cube_0.05.tet --input_filled_pc ../objects/filled_sphere.obj --input_surface_pc ../objects/surface_sphere.obj
2. torus: python main.py --name sphere --init_cube ../objects/cube_0.05.tet --input_filled_pc ../objects/filled_torus.obj --input_surface_pc ../objects/surface_torus.obj
3. G: python main.py --name sphere --init_cube ../objects/cube_0.05.tet --input_filled_pc ../objects/filled_g.obj --input_surface_pc ../objects/surface_g.obj

• Execute from src folder.

Results

Sphere	Torus	G