Voxelization & Volume¶

voxel_to_mesh: Convert meshes and point clouds to voxel grids. Perform volumetric operations like binary erosion and marching cubes reconstruction.


Methods	`voxelization`, `binary_erosion`
Input files	`.ply`, `.obj`, `.stl`, `.off`
Bundled datasets to try	`open3d_bunny`, `open3d_icp_demo`, `kitti_raw_2011_09_26_drive_0001`, `medshapenet`, `c3vd_endoscopy_splats`, `hf_objaverse_glb_shard`
CLI	`pardesline run voxel_to_mesh.<method> input.ext [--param value]`

How it works¶

Voxelization discretizes a 3D space into a regular grid of cubic cells (voxels). Each voxel stores a scalar value (occupancy, SDF, density). The inverse process extracts a surface mesh from this volumetric representation.

Marching Cubes is the standard iso-surface extraction algorithm. For each cube in the voxel grid, it classifies vertices as inside or outside the iso-surface based on their scalar values relative to an iso-value $\tau$.

Binary erosion shrinks a solid by removing boundary voxels using a structuring element.

Methods¶

voxelization¶

Voxelization converts a triangle mesh into a 3D occupancy grid. Open3D creates a VoxelGrid by testing which voxels intersect the mesh surface. The voxel size $v$ controls the resolution of the grid.

import pardesline as pl

obj = pl.load("input.ply")
result = obj.run("voxel_to_mesh.voxelization")  # catalog defaults
result.provenance[-1]  # the step, every parameter used, duration, version

Output: .ply. Script: python -m pardesline.ops.voxel_to_mesh.voxelization --input ... --output ....

Defaults validated on: generated meshes in tests/test_ops_smoke.py

Parameter	Type	Default	Range	What it does	Why this default, when to change it
`voxel_size`	float	`0.01`	0.002 to 0.05	Size of each voxel cube. Smaller = more detail.	0.01 in file units gives a 100-voxel grid across a unit-sized object: coarse enough to run in a second, fine enough to see shape; the range (0.002-0.05) assumes objects around a metre, so scale it for millimetre data.

binary_erosion¶

Binary erosion shrinks a solid by removing boundary voxels. For a volume $V$ and structuring element $B$: $$V \ominus B = \{x : B_x \subseteq V\}$$ The eroded volume is then reconstructed back to a mesh using marching cubes at iso-value 0.5.

import pardesline as pl

obj = pl.load("input.ply")
result = obj.run("voxel_to_mesh.binary_erosion")  # catalog defaults
result.provenance[-1]  # the step, every parameter used, duration, version

Output: .ply. Script: python -m pardesline.ops.voxel_to_mesh.binary_erosion --input ... --output ....