Can I use hOCR files as source for OCR, thereby skipping Tesseract in the process flow?

[bluebox-steven]

Hi all. I'm wanting to use another OCR engine for generating the text output, but without having to create a plugin for it. Therefore, is it possible to use hOCR files as the source and get OCRmyPDF to skip Tesseract (ie. Tesseract/OCRmyPDF assume the OCR is already done, so just use the hOCR file)? If so, how best to achieve that (either command line or programmatically). Thanks!

[jbarlow83]

Yes, there are APIs in ocrmypdf._api that provide a "pdf to hocr" and "hocr to ocr pdf" in separate steps. An open source PDF app asked for this feature so they could use it to implement intermediate hocr editing. I don't know if that every got implemented, which is why it remains a private API, since I wanted to make sure the intended consumer could use it in its current form. For the same reason, there's no command line interface; you have to script it. So if you obtain hocr from somewhere, you can get it rendered and applied.

There are plugins like OCRmyPDF-PaddleOCR if you just want to use a plugin.

OCRmyPDF is pretty tightly coupled to Tesseract even if you bypass it for OCR - it won't really work properly unless it's installed.

[bluebox-steven]

Thanks for the reply. I found the discussions in #453 and API code over here: https://github.com/ocrmypdf/OCRmyPDF/blob/main/src/ocrmypdf/api.py

I'll test it out and see whether it will work for my implementation.