shapevolve 0.1.4

A genetic algorithm to recreate artworks using simple shapes, with Python 3.

A genetic algorithm to recreate artworks using simple shapes, with Python 3.

Evan Zheng, July 2020.

Initially designed on Google Colab with Jupyter Notebooks. Module built with assistance from PyCharm.

View on PyPI.

Sample Results:

The Starry Night:

Mona Lisa:

Girl with a Pearl Earring:

The Great Wave off Kanagawa:

How to install and use:

Standalone executable: (Windows)

Head over to releases and download the latest executable, shapevolve.exe.

In the command line:

Change the directory using cd to where the executable is located.

Then, run shapevolve.exe path/to/image_file.png to run the program with an image as input.

You can also run shapevolve.exe -h to view additional options.

Module:

Install shapevolve from PyPI using the following command:

pip install shapevolve

Here is some sample code to demonstrate how to use the module.

from shapevolve.evolver import Evolver
from PIL import Image

evolver = Evolver(Image.open("path/to/image.png")) # Sets up the Evolver object.

genome = evolver.evolve() # Evolves the genome.

image = genome.render_scaled_image() # Gets a numpy array that represents the evolved image.

genome.save_genome("path/to/save_checkpoint.pkl") # Saves the genome for later use.

More sample code can be found in samples.py, here.

Here is a detailed API reference.

Libraries and APIs used:

Third-party libraries used:

• NumPy for numerical computation with matrices

• ColorThief for grabbing color palettes from images

• Scikit-Image for computing image simularity

• OpenCV for building images from circles

• Pillow for image preprocessing

• Matplotlib for image display

• PyInstaller for building an executable file.

Built-in libraries used:

• Pickle for checkpoint saves and loads.

• Argparse for the CLI interface.

License:

See LICENSE file.

Thanks:

Ahmed Khalf’s Circle-Evolution module provided a great deal of inspiration.