Experiments using genetic and hill climbing algorithms to reconstruct images using layers of semi-transparent triangles. This video shows an image comprised of about 4000 triangles at generation ~800,000. Each generation, one or more of the triangles in the image are mutated (color or shape changed slightly), and if the mutant image is closer to the original image than the parent, the mutant passes it's "genes" (triangles) onto the next generation. I'll be posting the full story behind this on Instructables soon.
Made in Processing, latest code on Github: github.com/amandaghassaei/Genetic-Images