# .font: font.text_2
2020-05-14 13:42 nesdori  
# .font: font.text_1
another art cave; this time marilyn monroe print by andy warhol. i expected that this image would be nice to be "bd-objectized" but i was surprised how dynamic and vivid the colors of optimized image actually are. i expected some kind of compromise for some shades of colors but it almost looks like that the preview image consists of single color blocks intead of multi-color bd objects. i had to exclude voodoo rockfords again but for the sake of consistensy the location of inbox is again located where a voodoo rockford object would have been optimal.  
  
this cave is easier and not that frustrating than mona lisa :) but the amount of required diamonds is quite high and some kind of strategy should be utilized again. I thought about making this cave even harder by placing a hidden magic wall somewhere to convert amoeba to diamonds but that would have been too nasty (even with a hint). after all, this is supposed to be an art cave series and not a frustration series :)  
  
this time no hint about the exit location but if you take a look on the bigger map, the location should be obvious :)

# .font: font.text_2
2020-05-14 16:44 Arno  
# .font: font.text_1
amazing art again  nesdori ! nice to see that your optimization method also works well with more colourful images.  
  
and thanks for the less-frustrating difficulty level! ;-) this cave is  well suitable for high score compo i think. :-)

# .font: font.text_2
2020-05-14 17:09 DarkStoorM  
# .font: font.text_1
That is really nice! I wonder how would it behave with abstract art :D

# .font: font.text_2
2020-05-14 18:34 Arno  
# .font: font.text_1
now that we have a new art-cave:  shredder , could you show your run of 'mona lisa'? after many tries, whatever strategy i apply, i always end up with too many fflies near mona's forehead. but mostly, i'm already killed before reaching that point... :-o ;-)

# .font: font.text_2
2020-05-14 18:53 shredder  
# .font: font.text_1
once again beautiful cave  nesdori ! great work.  
  
 Arno , ofcourse :)

# .font: font.text_2
2020-05-14 19:56 DarkStoorM  
# .font: font.text_1
This is quite hard, I always end up missing 30-40 diamonds.  
  
Gotta act quicker :D

# .font: font.text_2
2020-05-14 20:05 nesdori  
# .font: font.text_1
 darkstoorm , there's plenty of time to act (and no special luck with amoeba needed). i think you should re-think your general strategy :)

# .font: font.text_2
2020-05-14 22:16 dustin  
# .font: font.text_1
who needs paint? use boulder dash! ;d

# .font: font.text_2
2020-05-15 21:18 Arno  
# .font: font.text_1
 nesdori , your algorithm must be quite advanced. yesterday i did some small experiment myself. i tried to create a similar bd mosaic of the painting "girl with a pearl earring" by johannes vermeer. i used some free online mosaic tools and delivered the main picture and png images of the bd elements, where i did an educated guess of a suitable color scheme. all results, however, looked awful... perhaps a problem with that painting is that too much different colors are present, but i'm not sure.  
  
just curious: how much manual work is it for you to run your algorithm on a given picture? (not counting the work of recreating the cave at this site, testing playability, etc...)

# .font: font.text_2
2020-05-16 09:02 nesdori  
# .font: font.text_1
 Arno , you must be kidding! :D :D "girl with a pearl earring" was actually my 2nd art cave attempt after mona lisa :) the result was not that bad generally but the colorful highlights (red lips and eyes) were pale. because the lips and eyes are kind of the soul of the painting i didn't publish it. the optimized colors were actually close to mona lisa's. the girl's scarf was blue and other shades were white, brownish or light brownish.  
  
highlights where a certain color is used only locally is a problem because when you optimize the entire image, colors which are used most of course get more weight but highlights are often more important. for that reason i made a weight map in my program where you can set weights for each block separately. mona lisa and marilyn were generated using same weights for each block. the algorithm actually suggested that one set of colors is better for marilyn than the one i finally selected but it was because the background was a little bit better but marilyn's face color was worse. because the face of course is more important i had to overrule the computer's decision :)  
  
there's only a little manual work. the image must be scaled and after that i can run first optimization and within a minute i can see if the image is worth "bd-objectizing". then i run several optimizations from random starting points (or i can set an interesting starting point manually) to find out what is the global maximum (or close it) among hundreds of local maximums. i don't have to guess the colors at all. guessing them is surprisingly challenging and in addition to guessing proper colors you have to guess that which order of them is the best.

# .font: font.text_2
2020-05-16 09:12 nesdori  
# .font: font.text_1
hmm, maybe i'll try "girl with a pearl earring" again but now with a weight map, i didn't use it first time

# .font: font.text_2
2020-05-16 12:15 Arno  
# .font: font.text_1
would be cool, hope you get a nice result. thanks!

# .font: font.text_2
2020-05-17 09:26 Arno  
# .font: font.text_1
my run free! :-)

# .font: font.text_2
2020-05-17 11:10 nesdori  
# .font: font.text_1
 Arno , nice run! almost identical to my test runs. you could have got even more diamonds easily but i guess just solving was enough :) my best test was 770+ but i decided to set required amount to 750 because even to reach that you just can't proceed blindly but must utilize a certain strategy.