Genetic Image Approximation Engine

September 16, 20251 minute

A custom-built genetic algorithm that evolves populations of semi-transparent triangles to reconstruct target images with stunning artistic fidelity. Each genome encodes triangle vertices, RGBA channels, and z-indices, while a pluggable strategy architecture (selection, crossover, mutation) optimizes pixel-level fitness across generations—producing compact, human-legible “compressed” artwork from any source image.

Argentina Flag Image Approximation Example

Built for the Artificial Intelligence Systems course at Buenos Aires Institute of Technology (ITBA).

GitHub Repo
Python NumPy Pillow scikit-image matplotlib PyYAML pytest