Not going into any of the technical details...but does someone know what "artistic patterns" mean? Because frankly all i'm seeing is some kind of coloured interferance?
Not sure if I'm missing the point, but I just can't get impressed by this in the same way the "emulate an artist's style with neural networks" did.
I'm on a road trip at the moment, and the patterns in the gallery look similar to some of the paintings on the walls of a couple of motels I've stayed at. It was my first thought on seeing the gallery.
I think they mean something that not obviously generated by an algorithm, or that could plausibly be judged as hand drawn, or that looks similar to things that have been hand drawn.
If you found this interesting I recommend looking at Picbreeder (http://www.picbreeder.org/). Collaborative interactive evolution of pattern producing networks produced very interesting results.
Do I understand correctly? You have f:R^2 -> R^3, where the input is the image coordinate, the output is an RGB value, and f is a neural net with activations, weights, etc, chosen so that the image (literal and mathematical) of the map looks aesthetically pleasing?
The weights are random. No attempt has been made to make sure the generated image looks aesthetically pleasing. That's the surprising bit: a random neural network generates good-looking images.
> That's the surprising bit: a random neural network generates good-looking images.
Despite the randomness in the weights, there's structure and coherence imposed by the activation functions and the manner in which activations are combined (the network architecture). There's nothing inherently magical going on: take a rational function with random coefficients and do domain coloring [1], and you'll get something interesting, too.
I've been meaning to experiment with adding random patterns into my painting workflow, mainly as starting points. I find it's usually easier to have something other than a blank surface to start from. This looks like it might help, thanks!
Most of my friends who are artists will tell you: art is 3 pages dissertations to tell you how anything you produce is art.
Hence the «fontaine de Jouvence» or the monochrome of Soulage.
Met newly educated kids from university in Art they told me art is just about the public power sustaining the market of art for the benefits of those who sponsor art.
Hence, the real «artistic pattern» should also come with a dissertation and evaluation of its value on the market.
If a program can produce thousands of it per hour, even at .1$ and you can sell them, you still get rich and famous at the end.
I think it's poor taste to judge art by how much money someone else is willing to pay for it. It's like judging software by how much money it can to extract from its users, or by how many people know the name of its authors. Both measures ignore the actual value that is provided.
What defines art is whether or generates a visceral, human response. Is it evocative? Does it spark emotion or memory?
An ML algorithm can luck out and get this right, provided enough training data. But would it sustain if its audience knew there was no human element? I think the machine behind the curtain would throttle any emotional response.
As to this implementation, I have to agree it doesn't produce much interesting from the sample images. I haven't dug into the code to know if this is implementation related or a symptom of its input.
See more info: https://github.com/chancejs/chancejs/issues/193