EA Art: Neural Style Transfer Portraits

by Ben_West1 min read3rd Oct 2019No comments


EA Art & Fiction
Personal Blog

While trying to decorate my house, I became frustrated that art is usually about things which don’t seem very important to me. In an attempt to be the change that I want to see in the world, I have been playing with using neural style transfer to create art about EA heroes.

(If you want to test yourself, count how many of these people you can recognize from their pictures alone. For me, it is surprisingly few.)

Neural Style Transfer (NST)

The basic idea behind NST is: take two images and then use an artificial neural network to create a third image using the style from the first one and the content from the second. Here is an example, using a picture of Norman Borlaug and van Gogh's The Starry Night as inputs:

The algorithm was first published by Gatys et al. and my code is based on this tensorflow notebook. Many tutorials and introductions to NST exist online, so I will not rehash it here.

Scientific Heros

One series of pictures is based on Science Heroes' list of top 20 scientists by number of lives saved.

EAs in Blue

Another series is based on a blue theme. These were created using LunaPic. These are just people who came to my mind, and I have no particular objective reason why they are the most important ones to use pictures of.

Further Directions

I started an Instagram account (@happy.effective.altruism) which you are welcome to follow to see more of these. EA does not currently seem to leverage Instagram very heavily (ACE is maybe one exception) and I’m curious to explore what picture-based EA looks like.

I’m also curious to hear other suggestions for EA imagery!

Code can be found on my GitHub and should be easy to run with Jupyter.

EA Art & Fiction2
Personal Blog


New Comment