Home | Community | Message Board


This site includes paid links. Please support our sponsors.


Welcome to the Shroomery Message Board! You are experiencing a small sample of what the site has to offer. Please login or register to post messages and view our exclusive members-only content. You'll gain access to additional forums, file attachments, board customizations, encrypted private messages, and much more!

Jump to first unread post Pages: 1
Invisiblenooneman
Male

Registered: 04/24/09
Posts: 14,561
Loc: Utah
Artificial neural networks create psychedelic images by enhancing features on featureless images * 1
    #21825039 - 06/18/15 06:55 PM (8 years, 7 months ago)


Escher would be proud.

Here's a news article describing it:
Quote:


http://www.theguardian.com/technology/2015/jun/18/google-image-recognition-neural-network-androids-dream-electric-sheep

Google sets up feedback loop in its image recognition neural network - which looks for patterns in pictures - creating hallucinatory images of animals, buildings and landscapes which veer from beautiful to terrifying

Thursday 18 June 2015 07.57 EDT

What do machines dream of? New images released by Google give us one potential answer: hypnotic landscapes of buildings, fountains and bridges merging into one.

The pictures, which veer from beautiful to terrifying, were created by the company’s image recognition neural network, which has been “taught” to identify features such as buildings, animals and objects in photographs.

They were created by feeding a picture into the network, asking it to recognise a feature of it, and modify the picture to emphasise the feature it recognises. That modified picture is then fed back into the network, which is again tasked to recognise features and emphasise them, and so on. Eventually, the feedback loop modifies the picture beyond all recognition.

At a low level, the neural network might be tasked merely to detect the edges on an image. In that case, the picture becomes painterly, an effect that will be instantly familiar to anyone who has experience playing about with photoshop filters:

But if the neural network is tasked with finding a more complex feature – such as animals – in an image, it ends up generating a much more disturbing hallucination:

Ultimately, the software can even run on an image which is nothing more than random noise, generating features that are entirely of its own imagination.

Here’s what happens if you task a network focused on finding building features with finding and enhancing them in a featureless image:

The pictures are stunning, but they’re more than just for show. Neural networks are a common feature of machine learning: rather than explicitly programme a computer so that it knows how to recognise an image, the company feeds it images and lets it piece together the key features itself.

But that can result in software that is rather opaque. It’s difficult to know what features the software is examining, and which it has overlooked. For instance, asking the network to discover dumbbells in a picture of random noise reveals it thinks that a dumbbell has to have a muscular arm gripping it:

The solution might be to feed it more images of dumbbells sitting on the ground, until it understands that the arm isn’t an intrinsic part of the dumbbell.

“One of the challenges of neural networks is understanding what exactly goes on at each layer. We know that after training, each layer progressively extracts higher and higher-level features of the image, until the final layer essentially makes a decision on what the image shows. For example, the first layer may look for edges or corners. Intermediate layers interpret the basic features to look for overall shapes or components, such as a door or a leaf. The final few layers assemble those into complete interpretations – these neurons activate in response to very complex things such as entire buildings or trees,” explain the Google engineers on the company’s research blog.

“One way to visualise what goes on is to turn the network upside down and ask it to enhance an input image in such a way as to elicit a particular interpretation,” they add. “Say you want to know what sort of image would result in ‘banana’. Start with an image full of random noise, then gradually tweak the image towards what the neural net considers a banana.”

The image recognition software has already made it into consumer products. Google’s new photo service, Google Photos, features the option to search images with text: entering “dog”, for instance, will pull out every image Google can find which has a dog in it (and occasionally images with other quadrupedal mammals, as well).

So there you have it: Androids don’t just dream of electric sheep; they also dream of mesmerising, multicoloured landscapes.





Here's the original image gallery:
https://photos.google.com/share/AF1QipPX0SCl7OzWilt9LnuQliattX4OUCj_8EP65_cTVnBmS1jnYgsGQAieQUc1VQWdgQ?key=aVBxWjhwSzg2RjJWLWRuVFBBZEN1d205bUdEMnhB

Here's the original post on which the news article is based:
http://googleresearch.blogspot.co.uk/2015/06/inceptionism-going-deeper-into-neural.html


These images are so insanely psychedelic that it makes me wonder if psychedelic visuals are produced through the same phenomenon but in a real living neural network in the human brain.

Discuss psychedelically.


Edited by nooneman (06/18/15 06:57 PM)


Extras: Filter Print Post Top
OfflineReposadoXochipilli
Here, there, inbetween
Male User Gallery


Registered: 08/30/05
Posts: 7,501
Loc: Sand and sunshine
Last seen: 19 days, 22 hours
Re: Artificial neural networks create psychedelic images by enhancing features on featureless images [Re: nooneman]
    #21825164 - 06/18/15 07:29 PM (8 years, 7 months ago)

always thought the visuals were just how your brain was interpreting the visual stimuli, maybe i am just splitting hairs between being tasked to do something vs dosing and the whole trip. guess i am thinking of tripping vs painting while tripping.

sweet post btw, some of those images are impressive


--------------------


Edited by ReposadoXochipilli (06/18/15 07:29 PM)


Extras: Filter Print Post Top
Invisiblenooneman
Male

Registered: 04/24/09
Posts: 14,561
Loc: Utah
Re: Artificial neural networks create psychedelic images by enhancing features on featureless images [Re: ReposadoXochipilli]
    #21825383 - 06/18/15 08:30 PM (8 years, 7 months ago)

I've seen a lot of images like this on 2CE in particular. Makes me wonder. Plus, the images of deer or whatever being enhanced reminds me of the kind of exagerated directions, patterns, and features you'll see in fur and other fur like things on psychedelics.


Extras: Filter Print Post Top
Invisiblepsi
TOAST N' JAM
Male User Gallery

Registered: 09/05/99
Posts: 31,456
Loc: 613 Flag
Re: Artificial neural networks create psychedelic images by enhancing features on featureless images [Re: nooneman]
    #21827036 - 06/19/15 07:46 AM (8 years, 7 months ago)

That's pretty cool.


Extras: Filter Print Post Top
Jump to top Pages: 1


Similar ThreadsPosterViewsRepliesLast post
* Keyboards, synthesizers, etc...? Adamist 1,311 9 04/08/03 02:11 PM
by Middleman
* A Review of "The Psychedelic Experience" CD by Leary World Spirit 2,616 11 10/06/03 02:30 PM
by kindkesey
* Help the psychedelic community grow, recruit! psikooz 1,276 8 04/19/04 02:02 PM
by filthysock
* support swami - get your sig image here!
( 1 2 all )
Vvellum 3,709 32 12/14/04 10:55 AM
by Ripple
* Famous psychedelic & other drugs users
( 1 2 3 all )
mntlfngrs 14,709 45 07/06/09 07:00 AM
by entheogensmurf
* Jesus on Psychedelics Ravus 895 9 01/12/05 07:39 PM
by Dark_Star
* Space/Psychedelic Wallpaper? Twirling 3,089 8 09/11/04 10:49 AM
by SkorpivoMusterion
* Psychedelic Art kroum 1,208 6 04/07/04 09:59 AM
by mjshroomer

Extra information
You cannot start new topics / You cannot reply to topics
HTML is disabled / BBCode is enabled
Moderator: Entire Staff
536 topic views. 5 members, 59 guests and 49 web crawlers are browsing this forum.
[ Show Images Only | Sort by Score | Print Topic ]
Search this thread:

Copyright 1997-2024 Mind Media. Some rights reserved.

Generated in 0.022 seconds spending 0.008 seconds on 14 queries.