Awesome

Primitive3D

Primitive3D: Self-Pratice Library for pytorch-based 3D Data Processing.

Installation

export OptiX_INSTALL_DIR=$OptiX_INSTALL_DIR # (optional)
python setup.py develop

Examples

• Ray Casting

import torch
import prim3d

vertices: torch.Tensor # input vertices
faces: torch.Tensor # input faces
# create the ray caster
ray_caster = prim3d.create_raycaster(vertices, faces)

# setting ray origins and directions
origins: torch.Tensor
dirs: torch.Tensor

# output contained
depths: torch.Tensor
normals: torch.Tensor
primitive_ids: torch.Tensor

# conduct ray casting
ray_caster.invoke(origins, dirs, depths, normals, primitive_ids)

• Marching Cubes

# toy examples from the PyMCubes
python examples/sphere.py
# we extract the sdf of bunny via `mesh_to_sdf`
python examples/bunny_sdf.py
# example for extracting a sphere using marching tetrahedra
python examples/sphere_tetrahedra.py

TODO

• RayCasting based on Optix or CUDA BVH
• Marching Cubes(Mitigate from CuMCubes)
• Add RayCasting Examples
• SDF From Mesh
• Documents

Acknowledgement

• Open3D
• PyMesh
• instant-npg
• AnalyticMesh

ResourcesPermalink

• How to get started with OptiX 7

Please feel free to discuss :)