tprei / genetic-imaging

A genetic algorithm that reproduces an image from scratch!

Genetic Imaging

A genetic algorithm that reproduces an image from scratch!

But why?

• It's a simple project of mine that was elaborated in about a day (but I will continue it cause it's really fun). These already exist, and there are tons on the internet that you can check out that are probably better than mine.
• Yea it's fun.

Some examples

Since there are a few parameters that can be tweaked on, results may differ, and the bigger your population size and number of epochs, the longer it will take to generate the output image.

I believe this one took less than a minute.

These roughly took, respectively: 0.5s, 2s, 5s, 9 min, 7 years (had to google it lol).

These one was just for curiosity sake: it took like 8 hours or so.

You can experiment it yourself by changing the parameters and using your own images.

How to run? It's quite simple :).

1. You can clone this repository by doing:

https://github.com/thiagop-usp/Genetic-Imaging.git

2. After that, you will need some python3 libraries, so you can run:

pip3 install -r requirements.txt

3. Then make sure you're running it with Python 3.7 (f-strings are the best):

python3 genetic.py

Improvements yet to be done:

• Code is kind bad. It can be much better modularized.
• It's quite slow, some things can be changed when calculating the fitness.
• There are much better ways to reproduce the individuals (and some better ways to maintain good ones
• The idea of mutation is very important here, but I didn't use it.
• Make the progress and estimated time better to understand haha.
• Make it work to grayscale and RGB as well. As of right now it's only black and white, and if you tweak the code around it won't work.
• Video mode: it will create a timelapse.

Updates:

• Works with greyscale now. RGB still in progress.
• RGB and Rectangles are giving me lots of bugs, not working well yet =(.