[video]https://youtu.be/bNxVPE5BcVA[/video]
1440p wil basically avoid the compression
Never understood why it has such an obsession with dogs and castles? I guess its just the amount of input images over time have sorta biased the algorithm that way?
So basically video games need more dogs. lots more dogs. Will bring up at the next meeting.
[QUOTE=LoneWolf_Recon;52231356]Never understood why it has such an obsession with dogs and castles? I guess its just the amount of input images over time have sorta biased the algorithm that way?[/QUOTE]
deepdream has to be told by the user what to make the original image look more like, in this instance its "dogs" so it generates a horrifying dogscape.
[QUOTE=No_0ne;52231400]deepdream has to be told by the user what to make the original image look more like, in this instance its "dogs" so it generates a horrifying dogscape.[/QUOTE]
So 90% of all the public demonstrations of deepdream are just people recycling the same input of dogs, birds and castles?
Damn, this video makes me really want to do acid again.
[QUOTE=No_0ne;52231400]deepdream has to be told by the user what to make the original image look more like, in this instance its "dogs" so it generates a horrifying dogscape.[/QUOTE]
Can the user select anything to base the image processing around? Every single image I've seen has been dogs and animal eyes.
I'm terrified to know how long this took.
If I recall correctly, DeepDream is fairly slow to process even a single image. With all the frames of footage used here... :disgust:
This gives me anxiety for some reason.
[QUOTE=Gmod4ever;52231917]I'm terrified to know how long this took.
If I recall correctly, DeepDream is fairly slow to process even a single image. With all the frames of footage used here... :disgust:[/QUOTE]
[QUOTE]Recorded gameplay footage processed under DeepDream neural network frame by frame resulting in 45 hours of generation for 2700 images.[/QUOTE]
[QUOTE=Im Crimson;52231501]Can the user select anything to base the image processing around? Every single image I've seen has been dogs and animal eyes.[/QUOTE]
By what I know a model has to be trained on the thing you want to appear in the image. So you would have to input a bunch of pictures of what you want to show up and have it processed for however long.
Like LoneWolf said, most demos just have the stock animal model.
DeepDream is how I imagine cosmic horrors to look, what with them causing madness on sight.
Because otherwise the whole "madness on sight" is just bollocks. Like, would a modern day person go mad seeing Cthulu or would all the heavy metal album covers prove some immunity? Because we've all seen Cthulu and he's nothing compared to the fetish shit on Deviant Art.
Sorry to sidetrack things but I've had this thought for a while and this is the first DeepDream thread I've seen since...
[QUOTE=LoneWolf_Recon;52231407]So 90% of all the public demonstrations of deepdream are just people recycling the same input of dogs, birds and castles?[/QUOTE]
For me it has lost its charm by now due to that
Holy shit that was an acid trip and a half.
[IMG]http://i.imgur.com/bO1uFUH.png[/IMG]
More like dogdream
Sorry, you need to Log In to post a reply to this thread.
