About a month ago there was a Google research blog
запись about the visualization of a neural network - say, how she sees the image in the processing , Pestryaev pictures such as this.
People began to wonder how this can be done independently and in a couple of weeks, a new
запись with
source for githab , which enable all to do on their own.
But there were those who quickly became bored, so carefully rolled psychedelic.
This KDPV:
Take on the source githabe:
github.com/graphific/DeepDreamVideo
How does this pose and how to meet those or other dependencies can be found in the detailed manual on REDD:
www.reddit.com/r/deepdream/comments/3cawxb/what_are_deepdream_images_how_do_i_make_my_own
there is, as it all run.
Actually if you incite visualizer on static images - it's not so bad, really, the pictures begin to play with new colors
And people with new eyes:
But if incite visualizer with trained neural network to search for images of animals on Colorino video, then ... well, see for yourself the way:
While the authors are actively fixes bugs in the repository, and participate in the discussion, so join in the fun.
The link
www.youtube.com/playlist?list=PL1z-xdY3wUl_yWU_KK49QLCe9UAp7_S3R You can find the rest of the video, processed by #deepdream
Good luck Friday!
Source:
geektimes.ru/post/253240/