Compute the smallest bounding ball of a point cloud. Cython binding of the popular miniball utility. Fast!
Project description
cyminiball
A Python package to compute the smallest bounding ball of a point cloud in arbitrary dimensions. A Python/Cython binding of the popular miniball utility by Bernd Gärtner.
To my knowledge, this is currently the fastest implementation available in Python. For other implementations see:
miniballcpp
Python binding of the same C++ source (miniball)miniball
Pure Python implementation (slow)
Installation:
The package is available via pip.
python -m pip install cyminiball
Usage:
import cyminiball as miniball
import numpy as np
d = 2 # Number of dimensions
n = 10000 # Number of points
dt = np.float64 # Data type
points = np.random.randn(n, d)
points = points.astype(dt)
C, r2 = miniball.compute(points)
print("Center:", C)
print("Radius:", np.sqrt(r2))
Additional output can be generated using the details
flag and compute_max_chord()
.
C, r2, info = miniball.compute(points, details=True)
# Returns an info dict with the following keys:
# center: center
# radius: radius
# support: indices of the support points
# relative_error: error measure realtive to r2
# is_valid: numerical validity
# elapsed: time required
#
# The maximal chord is the longest line connecting any
# two of the support points. The following extends the
# info dict by the following keys:
# pts_max: point coordinates of the two points
# ids_max: ids of the two extreme points
# d_max: length of the maximal chord
(p1, p2), d_max = miniball.compute_max_chord(points, info=info)
See examples/examples.py for further usage examples
Build package
Building the package requires
- Python 3.x
- Cython
- numpy
First, download the project and set up the environment.
git clone "https://github.com/hirsch-lab/cyminiball.git"
cd cyminiball
python -m pip install -r "requirements.txt"
Then build and install the package. Run the tests/examples to verify the package.
./build_install.sh
python "tests/test_all.py"
python "examples/examples.py"
## Performance
For a comparison with miniballcpp
, run the below command. In my experiments, the Cython-optimized code ran 10-50 times faster, depending on the number of points and point dimensions.
python "examples/comparison.py"
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for cyminiball-2.1.0-cp38-cp38-macosx_10_9_x86_64.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | beb4a13e645324fc5cbe5b687ffebf68a8923e75288f5de8664ecd4e95996664 |
|
MD5 | 9bab2a0ffdb71416ce5865fc7c63a9e8 |
|
BLAKE2b-256 | 3298a854545b258331198cbd82b277cd5ede335a5cf3929be4ba587e38583382 |