Transcription about pero-ocr

[ikiss-fit]

Hi, if you are trying to process this document in the pero-ocr web application (pero-ocr.fit.vutbr.cz), you can use the Complex printed and handwritten layout model. For the baseline detector model, select Universal. For the OCR model you can definitely try Universal Transformer, but it will probably be quite slow. Otherwise, you can try Czech Fraktur Printed or German Fraktur Printed on this document, but without the language model (leave NONE selected). Of course, you can try different settings to see what works best for you.

If your question was not about the pero-ocr web application, then I would need more information to assist you.

[halaneji]

Thank you so much for your reply,
My question about the model when i run the test in my local repository, because I use the pero-ocr model that is shared.

With this script how can i select the Universal Transformer model please ??

[ikiss-fit]

The Universal Transformer model is not publicly available. It is available only in the pero-ocr web application (pero-ocr.fit.vutbr.cz).

[halaneji]

is it possible to share it with me please??

[ikiss-fit]

I'm sorry, but we will not share the model with you. If you want to process documents with our latest models, you can use the free web application pero-ocr (pero-ocr.fit.vutbr.cz). In case you need to process a really large amount of documents, we could discuss it and maybe the processing could be done via API, but that would probably be for a fee.

[halaneji]

Yes I will process a large amount of documents, for this reason I should have a solution for that.
Thanks for your response!

[ikiss-fit]

OK, I would suggest continuing the conversation via email. Could you send me an email to [email protected] so we can discuss this further?

[halaneji]

Yes for sure, thank you so much.