A fun image-processing project marginally related to my learning theory research (github.com) 36 points by NarcolepticFrog 10y ago ↗ HN
[–] aendruk 10y ago ↗ You should probably remove the 28-MB image from the readme. [–] NarcolepticFrog 10y ago ↗ Heh, good point. I'll do that when I get home tonight. Thanks! (Edit: done)
[–] NarcolepticFrog 10y ago ↗ Heh, good point. I'll do that when I get home tonight. Thanks! (Edit: done)
[–] johnloeber 10y ago ↗ This reminds me of http://codegolf.stackexchange.com/questions/50299/draw-an-im.... There were some great, rather creative results: in particular, using weighted poisson disc sampling or fuzzy k-means returned some really nice images. [–] NarcolepticFrog 10y ago ↗ You're, right, the idea of the challenge seems to be almost identical :) Thanks for the link - their images are awesome!
[–] NarcolepticFrog 10y ago ↗ You're, right, the idea of the challenge seems to be almost identical :) Thanks for the link - their images are awesome!
[–] krazydad 10y ago ↗ This is very similar, visually, to the "Crystallize" filter in Photoshop. Here's a sample of the Hubble image run through that:http://i.imgur.com/dVEzOZi.png [–] NarcolepticFrog 10y ago ↗ Yeah, it looks pretty similar! I guess that filter probably does something similar, except with uniformly sampled sites. Thanks!
[–] NarcolepticFrog 10y ago ↗ Yeah, it looks pretty similar! I guess that filter probably does something similar, except with uniformly sampled sites. Thanks!
[–] flashman 10y ago ↗ Here's a somewhat-similar tool that uses Delaunay triangulation: https://github.com/timbennett/delaunayThe frontend has a few issues and seems to break after running more than one or two images, if anyone wants to submit a patch.
7 comments
[ 198 ms ] story [ 320 ms ] threadhttp://i.imgur.com/dVEzOZi.png
The frontend has a few issues and seems to break after running more than one or two images, if anyone wants to submit a patch.