Building a Font converter


I am new to PerceptiLabs, but have managed to follow the guide and build my first tutorial model.

I would like to build a model that can read Fraktur (Ancient German Script) and convert it to modern western font.

How would one go about that?
I am aware what I need is lots of data, aka text in old font and modern font, for training and testing.
But how to go about organising that data, roughly?

Cause of course I can take pictures of single characters, label them a, b, c and so forth.
Or pictures of whole words, and then label those.

But is that really the best way to go about things?

I think not.

That’s why I ask :wink:

1 Like

Hi @Sternenfisch

Welcome! My day is just finishing but I saw your post and thought that’s a really fun idea.

I will try to share some thoughts (later) tomorrow, but one question 1st - you said you can “take pictures of single characters” - that might not be the most productive/efficient approach?

How would you feel about using Python to create single characters in a 28 x 28 pixel image - the code would also generate the labels and you could switch fonts to generate input and target font images directly?

I also think there should be better ways… till tomorrow

In the meantime - there’s no space for it in the profiles as far as I can see so everyone here’s a bit “fuzzy”… if you feel like telling us a little about yourself it would be a good way of getting to know each other… :slight_smile: