Radiohead is leading
category: general [glöplog]
http://createdigitalmusic.com/2008/07/14/cdmotion-radiohead-use-creative-commons-for-music-video-data-audiovisual-stems-the-next-big-thing
giving the raw assets of their current videoclip to encourage video remixes.
doesn't that inspire you for more demo remixes ? :)
btw about remix and hybridizing, i was wondering if any AI student or researcher has already tried to apply genetic algorithms to video generation and demos, as an experimental way to combine assets from different sources ? the problem would probably be to compute the fitness function that would give nice results, but it could be simple at the beginning then refined at each iteration...
giving the raw assets of their current videoclip to encourage video remixes.
doesn't that inspire you for more demo remixes ? :)
btw about remix and hybridizing, i was wondering if any AI student or researcher has already tried to apply genetic algorithms to video generation and demos, as an experimental way to combine assets from different sources ? the problem would probably be to compute the fitness function that would give nice results, but it could be simple at the beginning then refined at each iteration...
cant we just shoot radiohead into the sun first?
the short video of my realtime early attempt last night (41 megs to download, I binaryced and quantified the dataset).
/o_O\ far more accurate than the original wireframish radiohead video!
ah great, now my work in progress from the last 2 years will be credited to zest's mastermind.
radiohead is leaking
ps: it's all about doing it for real, don't worry, keops will remind me of that in the next lines ;)
reminds me of how the beastie boys came up with their av logo contest some years ago and made some of the demoscene take part.
dalezy: Awesome, I fucking shot that! was pretty sweet as well!
This is far from new, and Radiohead is a bunch of faggoty wank, but that's my personal opinion. :)
This is far from new, and Radiohead is a bunch of faggoty wank, but that's my personal opinion. :)
iq: good interpolation, did you use RBF?
radial basis functions? nop. I don't interpolate anything, just draw a gaussian sprite at every point of the pointcloud (it was a quick test).
ok, because in original video they used those small dots and your sprites cover bigger surface, so I thought those might be polygons.
BTW I tried the executable and I see black screen on ati2600mobility (vista)
BTW I tried the executable and I see black screen on ati2600mobility (vista)
what's a gaussian sprite ?
watch sprite called star4 :)
i wonder if 3D scanners will become the nightmare of demos like flatscanners were for the no-copy pixel scene :p
kraviz> what ?:/ star4 ?
bleh i still don't get it :(
bleh i still don't get it :(
quardian: download iq's early attempt, unpack it, go to "data" directory, open "star04.tga" - that's it.
And it looks like this not so random image:
And it looks like this not so random image:
oh
indeed i didn't download his attempt :) - only watched the video
thank you for detailing
indeed i didn't download his attempt :) - only watched the video
thank you for detailing
in fact that star04 sprite is not used, the gaussian one is generated, but it's similar yep.
yesterday i resampled the dataset into a regular grid (all the scan is just a heightmap, isn't it?), not only it compresses better (no need to store x and y) but also allows for easy polygonalization (terrain), point interpolation (morphing), normal estimation (light the face?), easy filtering, faster rendering and less memory usage. i can give the results if anyone is interested.
yesterday i resampled the dataset into a regular grid (all the scan is just a heightmap, isn't it?), not only it compresses better (no need to store x and y) but also allows for easy polygonalization (terrain), point interpolation (morphing), normal estimation (light the face?), easy filtering, faster rendering and less memory usage. i can give the results if anyone is interested.
iq i'm curious, what soft are you using to reorganize the plots, something like matlab or is excel enough ?
i always go for the simplest solution... i used C. 16 lines for parsing the file and projecting the points. also 25 lines for hole filling. it's simple but it works, and quality seems to be more than enough (images look sharper than the video i posted).
doh why did i ask! hope to see the result in a rgba prod :)
Are you going to release it at Euskal? :D
Nice work iq!
curses, it would have been fun to pull the data they have into my stuff and see how it came out, but I'm not going to write a csv importer for it when I have an xml one already there.
Anyone know of a free tool that will convert csv to xml in bulk?
Anyone know of a free tool that will convert csv to xml in bulk?