Artistic reproduction of images with geometric shapes.
Primipy is a command-line application that reads a reference image, and produces a set of geometric shapes that approximate the reference when drawn in sequence.
The algorithm adds one shape at a time. Each one goes through randomization and optimization steps. In detail:
- An arbitrary random number of shapes are tested, and the one that minimizes the error against the reference is chosen (randomization);
- The chosen shape is locally optimized via vertex mutations, and the best mutation is chosen (optimization).
Install Python 3.x, then
pip install -r requirements.txt
python main.py -i <your_reference_image> -o <output file> -n 200 -v
A SVG with vector shapes will be produced in the same directory of the output file.
| Option | Default | Description |
|---|---|---|
i |
n/a | input file |
o |
n/a | output file |
n |
n/a | number of shapes |
iters |
100 | number of iteration (randomization step) |
v |
False | verbose output |
vv |
False | more verbose output |
If using PNG as output format, Primipy stores some information in the final output, like how many shapes were used, and other useful data. Imagemagick's identify utility can be used to inspect this information:
identify -verbose <image.png>
This was inspired by the Primitive program by Michael Fogleman (https://github.com/fogleman/primitive/). I was fascinated by the beautiful results produced by such a randomized process and decided to explore it in more depth.
