The first exercise of Udacity's Deep Learning Nanodegree course asks the learner to use pretrained deep learning models to restyle photos according to famous paintings. During our recent behind the scenes tour of NASA's Johnson Space Center, I took this photo of one of the spacesuits, showing our reflections in the visor. I believe that this is a Russian spacesuit. American astronauts cross train to use Russian suits in case of an emergency.
The photo above is styled after Rain Princess by Leonid Afremov. This uses Logan Engstrom's Tensorflow-based Fast Style Transfer Convolutional Neural Net, along with a checkpoint that loads the trained parameters for the Rain Princess transform.
For comparison, here is the original photo that I took:
If you are interested in learning more about how to do this, check out Udacity's Deep Learning Nanodegree. The first couple of lessons are part of the free preview. You can also just download the style transfer code from the GitHub link above, but you will also need all of the ancillary packages to run it. The Udacity course walks you through setting this up. It is pretty simple if you have some experience with command lines and Unix-like environments.
A link to this post is on my twitter account:
https://twitter.com/johnsontoddr4/status/1019668489387347973
- johnsontoddr4
Proud member of #steemitbloggers
