A pipeline for turning images into paintings by Piet Mondrian. Interested in seeing examples? Check out the Twitter bot @PietMondrianAI and its respective repo mondrianify-twitter.
Using Python version 3.7, run pip install -r requirements.txt. Then run python MondrianPipeline.py. The script will draw a random photo from Unsplash and apply the transformation. The default directory output will be created and the image files will be placed inside. Similar to mondrianify-twitter, you can import this code by cloning this repository, placing it as a subdirectory, and running:
from mondrianify.MondrianPipeline import MondrianPipeline
random = True
mp = MondrianPipeline(image_path, random=random)
mp.apply_image_transform()The overarching class to help usher an image through the entire transformation. As it steps through the pipeline, it periodically saves the images output by the helper classes to a defined output directory. It relies on the classes in helpers to complete most phases of the process.
- BorderBuilder.py: Helps apply Holisticly-Nested Edge Detection to an image so that we can pull out its major features.
- ColorBuilder.py: Determines the colors used in a Mondrian painting. It draws from
colors.py, a file created by sampling from Mondrian's palette. - LineBuilder.py: Create many KMeans models to get a rough sketch of the segments that define an image. Then build out a Mondrian framework from those sketches.
- Painting.py: Combines the LineBuilder and ColorBuilder classes to create the final Mondrian painting.