ox-vox-nns 0.1.0

No project description provided

OxVoxNNS - Oxidised Voxelised Nearest Neighbour Search

A performant (for large numbers of query points) nearest neighbour search implemented in rust

Usage

Basic usage:

from ox_vox_nns.ox_vox_nns import OxVoxNNS

indices, distances = ox_vox_nns.OxVoxNNS(
    search_points,
    max_dist,
    voxel_size,
).find_neighbours(
    query_points,
    num_neighbours
)

TODO

PyPI

• Investigate building release binaries and pushing with pipx (manual)
• Investigate automated with cibuildwheel

Performance testing

• Create a function to generate test data, with the following parameters:
  • Number of search points
  • Range of point coordinate values (probably as a scalar, e.g. 15 => XYZ values in range [0, 15])
  • (Harder, do later) Distribution of points, i.e. some way of making the points less evenly spread (this can severely impact the performance of KDTrees)

Useful function: np.random.random((num_points, 3))

Plotting

• Compare KDTree implementations - run the same nearest neighbour search using different libraries and plot results (x-axis: num points, y-axis: processing time)
  • scipy.spatial.KDTree (native python)
  • sklearn.neighbours.KDTree (C++)
  • open3d.core.nns.NearestNeighbourSearch (multithreaded C++)
  • OxVoxNNS (rust)