What is the suggested approach for ocr pdf?

[nissansz]

What is the suggested approach for ocr pdf?

[JorjMcKie]

Can you please be more specific? Which tool to use? How to decide if at all to OCR? ...
You may want to read this as a start: https://medium.com/p/e455465acb03, also here: https://artifex.com/blog.
Another good source to contact others is PyMuPDF's Discord channel.

[nissansz]

Just convert image pdf to text bi-layer pdf with good layout

[JorjMcKie]

Try OCRmyPDF. If you want an integration within your PyMuPDF script use the integrated Tesseract access. There are example scripts in the utilities repo.

[nissansz]

Tesseract is hard to train and is low in accuracy. I want to change to paddleocr, any method?
Or change to other json based ocr result to format pdf finally.

[JorjMcKie]

Tesseract is hard to train and is low in accuracy. I want to change to paddleocr, any method? Or change to other json based ocr result to format pdf finally.

You obviously know better than me, so please communicate your findings in this channel.

[nissansz]

ok