He who knows how to appreciate colour relationships, the influence of one colour on another, their contrasts and dissonances, is promised an infinitely diverse imagery.
- Sonia Delaunay
Colour is my day-long obsession, joy and torment.
- Colour is my day-long obsession, joy and torment.
 

 

Wednesday, December 10, 2008

Genetic Programming: Evolution of Mona Lisa

Posted by Charley Parker at 9:11 pm

Genetic Programming: Evolution of Mona Lisa
Trial and error.

What artist has not at some point resorted to “I’ll just try this and see if it looks better.“?

You might say that, in light of Darwin’s model of natural selection, nature itself does the same: make a genetic mutation or two, or a billion, and see what works.

Swedish programmer Roger Alsing has created a playful experiment in “genetic programming” applied to image making, in which he wrote a small program for rendering 50 translucent polygons into an image area.

He set it to mutate slightly with each iteration, so that each pass of the program produces a different distribution of the polygons (the “genetic mutation”).

The fact that the polygons are translucent allows for many smaller subtle shapes within the composition, produced by overlapping areas of color, like laying an area of yellow glaze over both blue and green shapes in an oil painting.

At the end of each rendering sequence, the program uses a “fitness function”, basically a small routine to compare the resultant image pixel by pixel with a target image, in this case an image of the Mona Lisa.

Based on the “fitness” of the image, the program keeps either the new “dna” or the existing “dna”, whichever is more like the target, as the basis of the next mutation and iteration.

Trial and error. Survival of the fittest.

There is a selection of images on Alsing’s blog showing various renders, from which I’ve pulled a few representative samples, above. (For those who are programmatically inclined, there is also a faq with some of the basics.)

Under each of the sample images is a filename that shows the number of times the program had to run to reach that particular image.

The one at bottom-right shows 904,314 incidences of “I’ll just try this and see if it looks better“.

[Via Kottke]

Posted in: AmusementsDigital Art   |   5 Comments »

5 comments for Genetic Programming: Evolution of Mona Lisa »

RSS feed for comments on this post.

  1. Comment by JohanM
    Thursday, December 11, 2008 @ 3:24 am

    I love seeing experiments like this. I have merely peddled at a child’s level compared to the more advanced things being done by people far smarter and with more dedicated time on their hands.
    Always fun to see a fellow Swede do something fun too. ;)

  2. Comment by allan
    Friday, December 12, 2008 @ 8:59 pm

    wow this one seems nice and interesting looks like one of mine thanks.

  3. Comment by Michelle Reinke
    Monday, December 15, 2008 @ 10:22 am

    This is getting highly discussed right now – I love seeing new ways of combining the arts and sciences. The more conventional examples include architecture (engineering and art) and medical illustration (biology and art). Now there’s evolution and art! You have to wonder if someday computers will spontaneously spawn beautiful visual works.

  4. Comment by shahrzad
    Tuesday, December 16, 2008 @ 4:46 am

    Did you work with photpshope cut out filter ?I worked by it ,very beautiful

  5. Comment by Hanna
    Thursday, October 29, 2009 @ 1:44 pm

    I don’t know why but I really like the 2 pictures in the middle. I could definitaly hand up those up on my wall. I think that by using digital art you can spawn beatiful visual works in a very interesting way.

Leave a comment

(required)

(required but not published)

 
Display Ads on Lines and Colors (1st tier): $25/week or $75/month.

Please note that display ads for lines and colors are limited to arts related topics and may not be animated.
Display Ads on Lines and Colors (2nd tier): $20/week or $65/month.

Please note that display ads for lines and colors are limited to arts related topics and may not be animated.




Donate Life

The Gift of a Lifetime