This project is a simple but fun demonstration of a genetic algorithm applied to image generation. It starts with a population of random images and evolves them over generations to resemble a target image. The output is an animated GIF that shows the entire evolution process.
Interesting approach! The title is misleading though. I do not understand how this relates to pixel art. Pixel art is placing pixel on a constrained canvas and choosing colors on a limited palette which this does not seem to be. Maybe this could be considered blurring effect or image reconstruction/approximation?
How do you think your example would compare against a simple lowpass + downsampling?
I think for most complicated pixel art, "reimagining" the image is necessary to keep it visually appealing and discernible at low-res, like in the examples here:
Lowpass filtering with downsampling is definitely a more efficient way to simplify images compared to my approach. This project is an experimental demo of a genetic algorithm evolving random pixel patterns to approximate a target image.
I agree that true pixel art often requires artistic "reimagining" to stay visually appealing. Thanks for your feedback!
Pixel art isn't just bucketing a higher resolution image into a lower resolution. If you're going to use that term, you should probably have at least a basic understanding of the medium.
Thanks for pointing that out. You're absolutely right. What I'm doing here is definitely not pixel art in the traditional sense, and I now realize the term may have been misleading. My apologies for that.
This. I wasn't expecting much, having tried similar claiming algorithms and been disappointing. But even among those, the example image used here looks barely better than downscale and posterize
reply