Pages

Tuesday 7 February 2012

Image reconstruction using genetic algorithm



In this video, a genetic algorithm is running to evolve the image (on right) which consists of 50 triangles. The algorithm determines the position, the size and the color of each triangle each time, than the whole image is evaluated and the difference with the original one (on left) is plotted on the graph.

Technical details:
The genetic algorithm was executed for 14 hours, evolving around 250 000 Generations. The DNA size was 300 bytes (50 triangles each having 6 bytes: 2 for position, 3 for RGB, 1 for transparency), the picture was 60 000 pixels (200*300), the final state reached a state of 80% of similarity (12 500 different pixels).
The Picture in this video is for the Fondation Deutsch de la Meurthe, Cité Internationale Universitaire de Paris, France.
The music is "Symphony of Science".

Developed by Assaad Moawad.

1 comment: