How to set up 'classification' model for text in Italics or small/all-capital letters?

[ConcaveTrillion]

Hello,

I'd like to try and see if I can use this tooling to classify text as in "Italics".

Would this be a feasible thing to do?

I'm not clear that a KIE model would do the trick here. Is there anything that I could use to do this today?

CT

[felixdittrich92]

Hi @ConcaveTrillion 👋,

We don't have such a functionality yet.
There is a really similar open issue #868, but to be honest it's not really high on our priority list.

We have several classification models which could be trained for a binary classification in your case: https://mindee.github.io/doctr/modules/models.html#doctr-models-classification

But as already mentioned there is no integration - so you would need to modify the classification training script to your needs and later on write your own inference pipeline.

Best,
Felix

[ConcaveTrillion]

I'll take a look at the classification. At the moment my priority also is to get a well-trained model that works for my use case (providing an initial OCR of public domain books to a group of human proof-readers). And I'll star that issue so I can follow it in the future.