How to start AI - Githubissues

edissyum / opencapture

Open-Capture is the one and only 100% Open Source intelligent document capture solution

https://kutt.it/OpenCaptureWebSite

GNU General Public License v3.0

24 stars 11 forks source link

How to start AI #886

Closed martin-korf closed 1 year ago

martin-korf commented 1 year ago

Hello everyone, We just installed version 2.9.0 of Open-Capture. Is there any way of how to start the AI in it? I don't have any opinion to upload documents, so how to do it? I tried to search it in the GitBook, but the Franche language is not so familiar to me.

Thanks for help

nathan30 commented 1 year ago

Hi,

The document isn't ready yet in English, sorry for that. Moreover, we didn't have the time yet to write AI documentation

To train your model you have to:

Locate your docservers folder (by default /var/docservers/opencapture/CUSTOM_ID/)
Choose if you want an AI model for Verifier or Splitter module (regarding your name, it's probably verifier)
In the /var/docservers/opencapture/CUSTOM_ID/verifier/ai/train_data/ folder create subfolders with the name you want
In each folder you just created, you can now add PDF files
Then in the UI you'll see your folders and you can start training

Btw, if you see any errors in translation, do not hesitate to create pull request :)

martin-korf commented 1 year ago

Hi @nathan30 , thanks for your message :) the step manuals was usefully and I can now train the AI :)

I noticed two issued with translation - one is with AI directly:

and the second is in the General>Configurations> The description is not translated :)

Should I create a separated bug report for both of them? :)

Best Regards Martin

nathan30 commented 1 year ago

Thanks for your report. I'm waiting your PR :)

The main issue about the second translation error is that it's related to the data_fr.sql file (the default database data in the application). For now, we didn't have the time and funding to create a data_en.sql (if you want to do it, feel free ;) )