I'm using the python bindings of open3d to down sample a point cloud. The library offers two methods to do so using voxels: voxel_down_sample and voxel_down_sample_with_trace
While the first returns only a down sampled point cloud, the latter returns a tuple of the down sampled point cloud and a matrix.
Per the documentation, the matrix is supposed to be a mapping between points from the original point cloud to down sampled one. However, the matrices I get are always of size n-by-8, with nbeing the number of points in the down sampled version.
My question is: how do I interpret that matrix in order to retrieve the mapping between the original and down sampled point clouds?
