fast-surface-nets
A fast, chunk-friendly implementation of Naive Surface Nets on regular grids.
Surface Nets is an algorithm for extracting an isosurface mesh from a signed distance field sampled on a regular grid. It is nearly the same as Dual Contouring, but instead of using hermite (derivative) data to estimate surface points, Surface Nets will do a simpler form of interpolation (average) between points where the isosurface crosses voxel cube edges.
Benchmarks show that surface_nets generates about 20 million triangles per second on a single core of a 2.5 GHz Intel Core i7. This implementation achieves high performance by using small lookup tables and SIMD acceleration provided by glam when doing 3D floating point vector math. (Users are not required to use glam types in any API signatures.) To run the benchmarks yourself, cd bench/ && cargo bench.
High-quality surface normals are estimated by:
- calculating SDF derivatives using central differencing
- using bilinear interpolation of SDF derivatives along voxel cube edges
When working with sparse data sets, surface_nets can generate meshes for array chunks that fit together seamlessly. This works because faces are not generated on the positive boundaries of a chunk. One must only apply a translation of the mesh into proper world coordinates for the given chunk.
Example Code
use fast_surface_nets::ndshape::{ConstShape, ConstShape3u32};
use fast_surface_nets::{surface_nets, SurfaceNetsBuffer};
// A 16^3 chunk with 1-voxel boundary padding.
type ChunkShape = ConstShape3u32<18, 18, 18>;
// This chunk will cover just a single octant of a sphere SDF (radius 15).
let mut sdf = [1.0; ChunkShape::SIZE as usize];
for i in 0u32..ChunkShape::SIZE {
let [x, y, z] = ChunkShape::delinearize(i);
sdf[i as usize] = ((x * x + y * y + z * z) as f32).sqrt() - 15.0;
}
let mut buffer = SurfaceNetsBuffer::default();
surface_nets(&sdf, &ChunkShape {}, [0; 3], [17; 3], &mut buffer);
// Some triangles were generated.
assert!(!buffer.indices.is_empty());
License: MIT OR Apache-2.0
73 Dec 25, 2022
110 Dec 21, 2022
74 Dec 27, 2022
76 Oct 20, 2022
47 Sep 23, 2022
310 Dec 22, 2022
26 Dec 16, 2022
31 May 10, 2022
13 Dec 20, 2022
70 Nov 28, 2022
23 Dec 19, 2022
12 Dec 20, 2022
22 Dec 22, 2022
3 Jul 30, 2022
34 Dec 20, 2022
3.2k Jan 6, 2023
2.3k Jan 3, 2023
4 Sep 15, 2021
310 Dec 22, 2022
69 Sep 20, 2022
63 Jan 1, 2023
20 Nov 30, 2022
1 Nov 25, 2021
7 Jul 1, 2023
5 Sep 6, 2023
8 Jan 3, 2024
3 Nov 25, 2021
11 Nov 30, 2022