stark / urnn

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

urnn

A neural network to smartly create coherent terminal colorschemes based on an image, which might then be used as background.

converting pixel to raw data (for vompatti)

say we have a pixel with the rgb value of (255, 120, 70). to display this information in a format that the neural network can understand, we have to convert it to 3 floats. each consist of the color/255 (max value).

255/255 = 1.00000000

120/255 = 0.47058823

70/255 = 0.27450980

so there you have it. our pixel is represented by 1.00000000 0.47058823 0.27450980

Examples

Done with data that wasn't in the training set

####Normal Way (wallpaper -> colorscheme)

1 2 3 4 5

####Reverse Way (colorscheme -> wallpaper) (Still WIP to generate backgrounds)

5

With a background generated from this website

6

With a background generated from this website

Blog post explanation

TODO

  • Turn the input of the network between [-1, 1] so that the training is faster
  • Get more data for the training
  • Test multiple color extracters and parameters for the training
  • Build the network in the opposite direction, from colorscheme to wallpaper
  • A procedural wallpaper generation/fetcher based on colors (for when the network is used in the opposite way)
  • Easy wrappers for all the mini-tools written

About

License:Other


Languages

Language:Perl 34.0%Language:Go 28.7%Language:C 23.8%Language:C++ 9.4%Language:Shell 4.2%