Mandylion

Using Recurrent Neural Networks to generate Icons and Hieroglyphs. The icons are encoded in a machine learnable format.

Fork me on GitHub

The idea behind deep is best explained in the corresponding blog post. The code can be found on Github

The image below shows the results of running the RNN over a set of icons:

Yes, you might say, but can't you do this much better using stable diffusion? Of course you can? Hop over to your nearest image generator and go for it. The clever bit here though is that we use use an RNN. And yeah, it is a sign of the times when RNNs were cool and transformers not yet invented.