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
ResourcesPermalink
Please feel free to discuss :)