--unpaper-args and --clean-final

[sojusnik]

In the release notes it is stated that

The argument --clean-final now implies --clean.

In the documentation however the following commands are shown as an example:

ocrmypdf --clean --clean-final --unpaper-args '--layout double' input.pdf output.pdf ocrmypdf --clean --clean-final --unpaper-args '--layout double --no-noisefilter' input.pdf output.pdf

So are those examples outdated, or do I misinterpret the statement in the release notes, since in those examples --clean could be omitted?

Additionally, in another thread you wrote

I am thinking that I will change the behavior so that --unpaper-args implies --clean and --clean-final.

Do you still plan to implement that?

[jbarlow83]

--clean-final does imply --clean now. It was useless without.

--unpaper-args implies --clean but not --clean-final. As such the awkward parameters in the docs are correct for the moment.

--unpaper-args might have some practical uses with only --clean, if you disable mask repositioning and are only trying to improve the OCR quality, but I'm concerned that's a fairly narrow use case. It so happens to be the use case of --clean on its own.

[jbarlow83]

This is a really messy corner of the interface and I'm not too happy with how it currently works, I've decided. I'll have to see if I can come up with something better. Open to suggestions.

[githubnavigator]

I'm thrilled with the new features for unpaper, so I'm glad to see you're still moving ahead on this functionality. My main usage for this is to split two-up scanned pages into 2 separate pages, removing the black edges and centering/deskewing the text onto two new sheets of paper. (I had a problem with that, but I'll post it under a new case.) If there's a way to bake that functionality into ocrmypdf without needing to use unpaper-args, that would be convenient, as I've been struggling with the args.

[jbarlow83]

Currently thinking:

• A new argument --unpaper will do what --clean --clean-final does now (clean the presentation and OCR images)

• --clean --clean-final --unpaper-args will all be removed

• A new interface will allow plugins, making it possible to write a short script that has the effect of --unpaper-args. This avoids the messy/error prone nested arguments on the command line and is a lot more powerful. For example Pillow or OpenCV functions could also be used on images. Replicating the existing functionality of --unpaper-args would be a motivating example.

[sojusnik]

Sounds reasonable! I'm wondering how user-friendly the approach with the plugins will be: Does someone have to know how to code to use the unpaper parameters, as we have now with --unpaper-args, or will it be a more or less simple external file with those parameters in it?

[jbarlow83]

It would need to be an external Python script/module that calls unpaper.

[githubnavigator]

Oh. Well, if possible, if you could please leave the "--unpaper-args" option in there, I'd really appreciate it. A separate Python script would significantly increase the complexity for me.

I was able to achieve my desired goals using these commands:

mutool poster -x 2 -y 1 in.pdf out.pdf
ocrmypdf --force-ocr --unpaper-args '--layout single' --deskew --clean-final in.pdf out.pdf

[sojusnik]

Do I understand correctly that with the python script approach it would be finally possible to split pages with unpaper?

[jbarlow83]

@githubnavigator The idea would be that all of that functionality could be packaged in a short script: ocrmypdf-poster in.pdf out.pdf.

Generally I want to make it easier for people to customize things and attract more contributions.

I am both working on making ocrmypdf usable as a library and making the library extensible. That should make it easier to implement page splitting, but I don't think would be doable with plugins alone. As it currently looks the way to do that would be use to use components of ocrmypdf as a library in a custom command line driver.