Slashdot Mirror

← Back to Stories (view on slashdot.org)

Open Source Genetic Image Generation Software

Posted by timothy on from the evolving-a-picture dept.

maloi writes: "A couple of years ago, I wrote some software to mimic the genetic image generation software Karl Sims describes in his paper Artificial Evolution for Computer Graphics. After neglecting it for a long time, I finally got around to it again a few weeks ago, cleaned it up a bit, and released it under the GPL. You can now find it at Sourceforge. There's also a bit more info about it at Kaizolabs. I should mention that this is my first foray into open source development, so if there's anyone interested in both working on it and perhaps lending some guidance to me, it would be greatly appreciated!"

9 comments

  1. Why was this posted? by zpengo · · Score: 0, Offtopic
    Are we going to post something every time someone with a clever idea posts something to Source Forge and pimps it on Slashdot?

    Yes, this is a clever idea, but there are *thousands* of clever ideas on Source Forge that are generally neglected?

    I guess the arbitrary "news" detector of Slashdot has reared it's ugly head again. :)

    --


    Got Rhinos?
    1. Re:Why was this posted? by adamy · · Score: 2

      Because it is something worth noting. SOmebody decided that it was news for nerds/stuff that matters.

      Genetic Algorithms are a cool concept, and w/o a PhD in CS, something that is beyond most peoples understanding. There has been a lot of atalk about it, and now there is a sample project so hackers that want to take a look at source code can dig through it.

      I will agree that there are thousands of things on SourceForge that are neglected. Mostly they are neglected by the people that posted them there.

      --
      Open Source Identity Management: FreeIPA.org
    2. Re:Why was this posted? by Sleepyguy · · Score: 1

      Actually, all in all, genetic algorithims aren't all that hard to understand. Though they can be tricky to tune.

      I found this book to be quite useful:
      Practical Genetic Algorithms
      by Randy L. Haupt, Sue Ellen Haupt

      http://www.amazon.com/exec/obidos/ASIN/047118873 5/ qid=1017860863/sr=2-3/ref=sr_2_3/002-8797793-49784 62

      _

      --
      b
    3. Re:Why was this posted? by corezion · · Score: 1

      Actually I had slightly more than a high school education when I worked on the Neural Bot for QuakeII which uses genetic algorithms and artificial neural networks. I have to agree, why was this posted? And it's java... *vomit*!

      --
      "There is no Death. Only a change of worlds."
    4. Re:Why was this posted? by corezion · · Score: 1

      I suppose I shouldn't just bitch and moan about it being in java. This could make a very cool GIMP plugin. That's my suggestion.

      --
      "There is no Death. Only a change of worlds."
    5. Re:Why was this posted? by vivian · · Score: 1

      I reckon it's definitely worth posting.
      This has finally brought me out of 3 years of lurking on Slashdot. After having a go at writing a genetic algorithm in perl that could pump out sort functions, I quickly became dissilusiond that relatively simply Genetic algorithms could be used for anything of interest besides generating functions to fit curves etc - or at least was beyond my immediate skills. (I usually develop boring back office/corporate stuff for a living involving databases etc.)

      This has definitely piqued my interest & the idea of parameterising IFS fractals & mutating them is a winner. Question: If you were using something like this as an extension to the Gimp, once you have a shape pretty close to what you want, could you stretch it a bit/modify part of it and have the "genetic" code adapt those changes before going on to new and more interesting mutations? giving evolution a push in a particular direction with relatively untalented artwork?

      So I draw, say a rough looking beastie for my lastest game, (having almost zero talent at such) have it mutate into something genuinely interesting and original, then touch it up and repeat as required? That truly would be software worth having/working on!

  2. Documentation, please. by Rick+the+Red · · Score: 2
    If you want our feedback on your code, don't require us to decypher it. Please document your code so we can understand it. Thanks.

    --
    If all this should have a reason, we would be the last to know.
  3. A similar idea, with implementation. by mahlen · · Score: 1

    I wrote something like this two years ago (though i was unaware of Sims's paper) based on the image generation ideas from this 1999 Slashdot piece, referring to the apparently still active Gallery of Random Art. My stuff allows you pick a size, though, so you can make desktop backgrounds :) You act as the environment, selecting images that you would like to see mutated. The mutations are frequently fairly subtle, so it can take several iterations. Sadly, there is no cross-breeding yet.

    With some trepidation, I'll point to my server which is running this. Be gentle, it's just one box and it's on a cheap DSL line (128Kb upload). You can see and download some examples of what it's produced on my server and also on WebShots here and here.

    On the odd chance that you'd like to see my code, email me. I don't have the ego to presume that this stuff is so interesting to people to have published it myself.

    mahlen

    In Riemann, Hilbert or in Banach space
    Let superscripts and subscripts go their ways.
    Our asymptotes no longer out of phase,
    We shall encounter, counting, face to face.
    --Stanislaw Lem, translated by Michael Kandel, "The Cyberiad"