//=time() ?>
Those wrong brush strokes might be realllyyy hard to get out
Color still has a long way to go. Moving to painting with color from black and white is super complicated, even for humans. I dont know what to expect from my algorithm.
The issue is that small errors in colors are so visible in the final image. I might hit a dead end here
Details are slowly looking good. Today I did a bunch of tests and decided to support a pass where it focuses on hand given masks. Non of these examples have it, but as you can see the eyes never have enough details.
Almost done with the black and white side of painting. Still some minor ideas I want to try out. Main thing would be color, there I still need better fitness function
A fitness function needs to take that in to account, a f(h) which is less sensitive to change of hue in green areas and more in the red areas.
Alternative I can distort the HSV space so that it resembles an artists color wheel and measure the euclidean distance there.
Break through! my simple painting AI can now (try to) paint more complicated shapes. Is genetic algorithm considered AI? I am not sure what is AI these days.