Add HOCR output as a sidecar option

[parkerhancock]

I have many applications where the physical location of text on a page is significant, and an existing codebase built around the HOCR html format.

What would make this library completely killer is an option to produce a sidecar file of the hocr data from Tesseract. I know that Tesseract natively can produce HOCR data, so the change shouldn't be difficult. The only question is how to integrate that into the existing command line interface.

Maybe a new option for --sidecar-hocr?

Flipping through the codebase now to see if there's an easy option.

[jbarlow83]

ocrmypdf has three PDF renderers.

One of them is called the hocr renderer and uses HOCR as an intermediate format. For your use case it might make the most sense to use the older hocr renderer, since you intend to hocr for other things.

So

ocrmypdf -k --pdf-renderer hocr

which will output a temporary folder with all working files, including the hocr files per page. The main drawback of the hocr renderer is that its support for non-Latin script is poor.

If you'd prefer to force generation of hocr files using the new (and default) sandwich renderer (best PDF quality, requires Tesseract 3.05.01 or newer):

ocrmypdf -k --tesseract-config hocr <rest of your arguments>

I will think about adding an option for hocr sidecars that involves less hackery, but this should do it for now.

[andrewjfreyer]

Any further thoughts on adding additional sidecar features?

[zweissman]

Your suggestion to use ocrmypdf -k --tesseract-config hocr <rest of your arguments> works great along with the keep-temporary-files=true. The only issue that I am currently having is where to find the temp files. The path is output while the ocr is running, but that temp path changes from run to run. Is there any way to query the ocrmypdf object to get the temp path so I know where to look for the .hocr file?

[hendursaga]

This probably is related, but I was wondering if there's a good way to store the hOCR on the side and then "apply" it to the PDF when needed. I want to retain the original PDF files without having to basically duplicate them, more than doubling storage costs.

[jbarlow83]

You can use the API functions in ocrmypdf.api to save a hocr and apply it later.

[hendursaga]

@jbarlow83 Which API functions? Are there command-line flag(s) that could do the trick just yet?

[jbarlow83]

OCRmyPDF/src/ocrmypdf/api.py

Line 383 in 3a75b20

def _pdf_to_hocr( # noqa: D417