Which marching cubes algorithm does this implement?

[glebarez]

marching_cubes from scikit-image provides 2 implementation for marching cubes algo:

• The algorithm [1]_ is an improved version of Chernyaev's Marching
  Cubes 33 algorithm. It is an efficient algorithm that relies on
  heavy use of lookup tables to handle the many different cases,
  keeping the algorithm relatively easy. This implementation is
  written in Cython, ported from Lewiner's C++ implementation.
• The classic Lorensen's algo

Which one does this library implements?

[schlegelp]

I'm not familiar with the details of either of the scikit-image implementations to be honest.

I wrote this package based on an idea I had for getting a mesh from sparse voxel matrices and didn't bother looking at existing algorithms as I figured they were all predicated on a dense matrix as input.

Ultimately, it is just using some numpy magic to find surface voxels in the sparse matrix and turns them into mesh vertices + faces. There is no handling of the different cases for half-edges you find in typical marching cube algorithms.