Sort of off topic, but not really. There was a paper recently presented at Siggraph that details the use of neural network computing to create color images from black and white source images. The network was trained on a dataset called "Places" and the authors indicate that the training data is best used to interpret "natural outdoor images." I figured I would give it a try using @dagoimaging's black and white interpretation of Gene's image to see what the model produces as output.
Attached is the result simply by copying @dagoimaging's posted black and white, converting it to grayscale and slightly reducing the contrast, downsizing it and running it through the routine published by the authors of the paper. Not too bad, all things considered.
kirk
Here is more information about the research:
http://hi.cs.waseda.ac.jp …projects/colorization/en/
and the publicly available code to run the model:
https://github.com …siggraph2016_colorization
and the citation for the code:
author = {Satoshi Iizuka and Edgar Simo-Serra and Hiroshi Ishikawa},
title = {{Let there be Color!: Joint End-to-end Learning of Global and Local Image Priors for Automatic Image Colorization with Simultaneous Classification}},
journal = "ACM Transactions on Graphics (Proc. of SIGGRAPH 2016)",
year = 2016,
volume = 35,
number = 4,
}
THIS IS A LOW QUALITY PREVIEW. Please log in to see the good quality stuff.
)
