https://github.com/numberoverzero/kmeans

python wrapper for a basic c implementation of the k-means algorithm
https://github.com/numberoverzero/kmeans

Last synced: 17 days ago
JSON representation

python wrapper for a basic c implementation of the k-means algorithm

• Host: GitHub
• URL: https://github.com/numberoverzero/kmeans
• Owner: numberoverzero
• License: mit
• Created: 2013-08-24T21:02:33.000Z (about 11 years ago)
• Default Branch: master
• Last Pushed: 2015-03-26T16:36:40.000Z (over 9 years ago)
• Last Synced: 2024-10-07T09:06:28.077Z (about 1 month ago)
• Language: Python
• Homepage:
• Size: 301 KB
• Stars: 24
• Watchers: 3
• Forks: 4
• Open Issues: 1
• Metadata Files:
  • Readme: README.rst
  • License: LICENSE

Awesome Lists containing this project

README

        
kmeans

===================

.. image:: https://travis-ci.org/numberoverzero/kmeans.png?branch=master

   :target: https://travis-ci.org/numberoverzero/kmeans

python wrapper for a basic c implementation of the k-means algorithm.

**Please review the limitations before using in any capacity where strict

accuracy is required.  There is no overflow detection, and negatives are not

supported.  tuple values cannot exceed 255.**

Installation

===================

::

    pip install kmeans

Usage

===================

::

    import kmeans

    means = kmeans.kmeans(points, k)

``points`` should be a list of tuples of the form ``(data, weight)`` where

``data`` is a list with length 3.

For example, finding four mean colors for a group of pixels::

    pixels = [

        [(15, 20, 25), 1],  # [(r,g,b), count]

        [(17, 31, 92), 5],

        # ... Lots more ...

    ]

    centers = kmeans.kmeans(pixels, 4)

In this case, the weights passed in may be the frequency of the pixels occuring

in an image, or some preference to pull the means towards a color.

Limitations

===================

All values must be **non-negative** **integers**, with the following

restrictions::

    r, g, b        [0, 255]        (uint8_t)

    count          [0, 4294967295] (uint32_t)

    maximum points 4294967296      (uint32_t)

    maximum means  256             (uint8_t)

    max iterations 65536           (uint16_t)

    max tolerance  65536           (uint16_t)

Inspiration

===================

http://charlesleifer.com/blog/using-python-to-generate-awesome-linux-desktop-themes/

I wanted to apply the implementation there to images much larger than 200x200.

Running a 4k x 3k image was approaching 60 seconds on a nice computer, so I

decided to rewrite the kmeans implementation in c.

Contributors

===================

* chinatsu_ - 32bit compatibility

.. _chinatsu: https://github.com/chinatsu