I think the algorithm lets you train it from your own images - the default adds a lot of lines, streaks and trails but I'm guessing this guy added some dog photos in for fun.

Imagine if they trained it up with spider photos.

From the description on the original site it seems like he just used Google's sample training set, which is mostly trained with dogs and snails or something. For some reason.

Now I want to train a version of it with all the standard Illuminati/conspiracy symbols to see what the world looks like through the eyes of a hardcore conspiracy theorist.

There's a web interface for it
http://psychic-vr-lab.com/deepdream/
But there's a big backlog of photos going through. Could take awhile.

Someone put a Hubble shot through...
http://psychic-vr-lab.com/deepdream/pics/12729.html

I've got to say, I don't find the stills NEARLY as interesting as the video. The stills look like someone took a The Mountain shirt and ran it through the "plastic wrap" filter in Paint Shop Pro.

There were stills before, but I'd love to see a Cyriak video through this.

Exactly, I sometimes see people as Dogs.

:D

You make this place worth visiting, Boomer.

glad youre still around Boomer!

Me too!