Genetic Programming: Evolution of Mona Lisa

genetic-programmingevolutionary-algorithmscomputational-artoptimization

Abstraction: Evolving Mona Lisa replica using 50 semi-transparent polygons

Key points:

Program maintains a DNA string encoding polygon parameters; each generation copies and slightly mutates the DNA, renders it to canvas, and keeps the mutation only if it reduces pixel-level distance to the source image
Challenge: approximate the Mona Lisa using only 50 semi-transparent polygons
Selection criterion is purely fitness-based (closeness to target image pixel comparison); no crossover, a simple (1+1) evolution strategy
Images at various generation counts show progressive convergence from noise toward recognizable portrait features
Demonstrates that hill-climbing mutation with a simple fitness function can produce visually compelling results over many generations

Connections: Genetic Programming · Evolutionary Algorithms · Computational Art