Dave Horner's Website - Yet another perspective on things...
Home
88 guests
Rough Hits : 3792032
moon and stars
how did u find my site?





 
our entire universe; minuscule spec in gigantic multiverse which is mostly lethal.



 
The man who grasps principles can successfully choose his own methods. The man who tries methods, ignoring principles, is sure to have trouble.
--Ralph Waldo Emerson
\begin{bmatrix} 1 & 0 & \ldots & 0 \\ 0 & 1 & 0 & \vdots \\ \vdots & 0 & \ddots & 0\\ 0 & \ldots & 0 & 1_{n} \end{bmatrix}

android dream of a Gordian Knot?

Friday, 05 October 2018 01:47
gordian knot dream
What's an android dream of a Gordian Knot?
¿Qué es un sueño androide de un nudo gordiano?
Qu'est-ce qu'un rêve androïde d'un nœud gordien?

DeepDream; a convolutional neural net which gives rise to algorithmic pareidolia, thus creating a dream-like hallucinogenic appearance.

Pareidolia (/pærɪˈdoʊliə/ parr-i-DOH-lee-ə) is a psychological phenomenon in which the mind responds to a stimulus, usually an image or a sound, by perceiving a familiar pattern where none exists.

anyway, there's this deepdreamgenerator website that essentially packages up the DeepDream algorithms for anyone to use. It's limiting in some of the settings and outputs; but it's free and easy to use. You can't get on Pouff level but it's cool. https://deepdreamgenerator.com/u/502959

Top left of the montage image is the original image. Images to the right and bottom are all generated from the original image and a style image. (you can see the styles and settings I selected from my profile) I used ImageJ to create the montage and to stack the images and average all 4 images together.

The montage processed images are done using DeepStyle which is basically asking the computer to dream us something that looks similar to the original and style image. Then taking the average of all 4 gives us something that looks very much like the original with the average providing us a mixture of the essence of the style images projected on the original image.

When I first started in on this, I was wanting to create a checkerboard mask that would let through portions of each of the different deepstyled images. Or to draw and cuts masks in the vertical or diagonal... this wasn't as easy as I would have liked.

A big part of what makes this "Deep" convolutional network idea so interesting is how you get more interesting results as you feed the results back into the machine to dream more on it.

The last image is created by taking the average image with the montage image as a style image. Which...to me....is so beautiful, really one of the coolest images I've ever produced. My android dreamt me my checkerboard/vertical/horizontal mask, but it didn't use hard lines, fuzzy about where one dream begins and the other ends.

Deep styling is like a lucid dream; guided by the essence of the original and style. Where are you guiding your android in it's dream? In the case of DeepDream, you're asking your android to obsess about it's models...whatever knot that might be.
What's this "Pouff level" level you speak of?!

The computer's hallucinating and it sees dog face.
gordian montage gordian average
  Next >
Last Updated on Friday, 05 October 2018 02:03