File Uploader and OCR

mzguy · 2021-01-28T17:13:29.206Z

Haven't implemented this but I have considered it. My thought would be to loop through notes and run the script (which will extract embedded PDF text) and then when it's done with the OCR, add an ocr tag to the note to indicate it's been done. I don't have the bandwidth to add something like this right now, but would certainly welcome a pull request! But yes, as a workaround you could extract the PDFs and let rest-uploader bring them in, create a preview image, and extract the existing OCR from the PDF.

For all the Evernote users out there that are jumping ship (including me!), this would be very welcome. Your script is working well for PDFs and images as they are added, but both I and many others have a large archive of notes that pre-date our use of Joplin.

If you or anyone can do this, I have a suggestion: provide a flag to go through all notes and process PDF files, or all notes with a particular tag. That way, a user can tag some notes with images that need OCR, and the script doesn't OCR every image, some of which won't need it.

Any chance this is on your near or medium-term radar?

kellerjustin · 2021-02-05T14:59:14.512Z

Medium-term at best. I can certainly see the benefit, but I just don't see myself having time to implement this anytime soon. Sorry.

mzguy · 2021-03-07T23:15:34.340Z

I just started with Joplin Linux, and installed via pip. I got an error:
ERROR: img-processor 0.11.0 has requirement Pillow>=8.1.0, but you'll have pillow 7.0.0 which is incompatible.

Rest-uploader seems to be working fine, calling with Python3 as a module. Should I be concerned something is suboptimal due to the error above?

Also, I have a couple of requests for enhancements or bug fixes:

I sent in a JPG that was approximately 4000x3000. Rest-uploader processed it fine, but it was downsampled to about 2000x1500 in the process. Is there some way to avoid or control the downsampling?

Also, Tesseract supports multithreaded processing. However, I'm not sure how to get that to be used within the rest-uploader workflow. Is it something simple I can activate?

mzguy · 2021-03-08T23:35:04.883Z

mzguy:

Also, Tesseract supports multithreaded processing. However, I'm not sure how to get that to be used within the rest-uploader workflow. Is it something simple I can activate?

I suppose I have the same question for pdf2image. Both that and tesseract benefit from multithreading, but I'm thinking the relevant variables / flags aren't being used since my CPU doesn't get much above 20% utilization.

kellerjustin · 2021-03-09T02:09:08.902Z

I had made some changes to the img-processor package a few weeks ago and must have pinned newer versions of things. Best practice would be to create and use a virtual environment, but the older version of those libraries should suffice. Thanks.

mzguy · 2021-03-09T18:42:22.786Z

Thanks. I'm a newbie here and learning. I'll try to get it set up through virtualenv, but to get everything with python back to Ubuntu stock, can I just pip3 uninstall rest-uploader, and all its dependencies one-by-one? I've not installed anything else with pip3 other than rest-uploader.

mzguy · 2021-03-09T19:41:23.305Z

I got it figured out, so never mind the question. Thanks for the venv suggestion.

myfta · 2021-04-13T08:16:19.579Z

Not sure if this is the best place to post this. Whilst I was catching up with how to use Tesseract with Joplin the two posts illustrate the importance of optimising the image before trying to OCR. If not done so already, maybe these ideas need to be carried through to the plugin developments?

https://levelup.gitconnected.com/a-beginners-guide-to-tesseract-ocr-using-pytesseract-23036f5b2211

Medium – 13 Jan 20

How to Use Tesseract on Windows

Tesseract is an optical character recognition software which developed by Google. Its an open source OCR tool. There are many versions of…

Reading time: 4 min read

mzguy · 2021-04-13T20:27:14.491Z

I'm getting frequent cases of attachments that don't have any text recognized, even when the text is already there from running Tesseract on the file! I can send an example if anyone can debug.

manouchk · 2021-06-11T17:56:13.664Z

THis plugin is not yet available in the official plugins list? I search for OCR and uploader in joplin and the plugin was not found.

JackGruber · 2021-06-11T18:12:25.876Z

This is no plugin for Joplin, this is a external script.

manouchk · 2021-06-12T11:16:10.244Z

Thank you. I have a strange error when trying to configure token:

/usr/lib/python3.9/site-packages/rest_uploader/api_token.py in get_token()
      8     else:
      9         token = input("Paste your Joplin API Token:")
---> 10         with open(".api_token.txt", "w") as f:
     11             f.write(token.rstrip())
     12     return token

PermissionError: [Errno 13] Permission denied: '.api_token.txt'

I opened the script in an interactive way in ipython and gave me '/usr/lib/python3.9/site-packages/rest_uploader' as the current path. I don't if there is an expected directory which is or could be define so that rest_uploader would find the .api_token.txt. Should that be the directory were files are uploaded (this is what makes sense to me)?

I installed the script in archlinux with the following command:

pypi2pkgbuild.py git+https://github.com/kellerjustin/rest-uploader

which constructs automagically an archlinux package from github. Don't know if this could have modified something in the script.

The script was installed in /usr/lib/python3.9/site-packages/rest_uploader/

 $ ls /usr/lib/python3.9/site-packages/rest_uploader. The following files were installed:

api_token.py  cli.py  __init__.py  __pycache__  rest_uploader.py

kellerjustin · 2021-06-14T13:47:19.181Z

I should refactor how that api token file is stored and put it in user space. You should be able to get around this by altering the permissions on the site-packages/rest_uploader folder or running the script as root the first time to create the .api_token.txt file.
Thanks for reporting!

harun27 · 2021-07-19T13:35:46.128Z

Is it possible to process all images and pdfs that are already in my notes?

I also always get this error:
Too few characters. Skipping this page Error during processing.
Am I doing something wrong or are the images just bad?

kellerjustin · 2021-07-19T14:05:08.418Z

rest_uploader only handles the creation of new notes from files. You might want to check out OCR Joplin Notes for what you're trying to do.
Thanks!

laurent · 2021-07-25T20:05:24.676Z

If the Tesseract lib reliable enough in general? I've read that it's difficult to use, but it seems you managed to make it work well?

kellerjustin · 2021-07-26T14:49:13.685Z

If it was terribly difficult to use, I wouldn't have been able to make it work !
I'll put it this way - the creator(s) of tesseract did the heavy lifting and made a pretty solid opensource OCR framework, and subsequently pytesseract made it easy to access from Python. I just had to hook into it. I've found it's much easier to deal with tesseract in Linux because package management makes it easy - in Windows, unless someone has released a better installer in the last six months or so, it's sort of a headache to install and involves setting environment variables.
Thanks!

laurent · 2021-07-26T15:04:51.176Z

Thanks for the info, it seems indeed easier than I thought. Does it successfully OCR documents most of the time? Are there any issues when the documents are not in English?

kellerjustin · 2021-07-26T15:17:14.569Z

It's been pretty solid. My handwriting is garbage so I don't expect it will work with any of my handwritten notes, but it seems to do well when processing a scanned text document. You pass tesseract a language parameter and other than the occasional testing, I only ever pass in --eng so I can't adequately answer your question as to how well it works with other languages. Hope that helps!

Chamo · 2021-12-05T22:13:50.454Z

I'm getting the following error message:
Check Tesseract OCR Configuration
UZN file C:\Users[username]\AppData\Local\Temp\tess_tvjq976a loaded. Estimating resolution as 258 UZN file C:\Users[username]\AppData\Local\Temp\tess_tvjq976a loaded. Warning. Invalid resolution 0 dpi. Using 70 instead. Too few characters. Skipping this page Error during processing.

Any ideas? From Warning. Invalid resolution 0 dpi. Using 70 instead. · Issue #1702 · tesseract-ocr/tesseract · GitHub it sounded like it might be an issue related to the metadata ("It means your image does not contain a resolution info in its metadata, so Tesseract warns you about this issue in the image and it tries to estimate the resolution by itself.") but I wasn't sure what to try to do to fix it.