OCR Poor results

Discussion for the End User use of OCR in PDF-XChange Editor and Viewer

Moderators: Tracker Support, TrackerSupp-Daniel, Sean - Tracker, Paul - Tracker Supp, Vasyl-Tracker Dev Team, Chris - Tracker Supp, Ivan - Tracker Software, Tracker Supp-Stefan

Post Reply
Crookie
User
Posts: 3
Joined: Thu Jun 06, 2024 2:01 pm

OCR Poor results

Post by Crookie »

I am getting quite poor results from OCR.
I want to convert a scanned PDF document into an editable PDF, without changing anything at all, just have an editable reproduction, but the results are consistently inconsistent, with any hand written notes being converted to garbage.
If anyone has any ideas I would be most grateful, I have 50 installs waiting on the back of this.
Top image is original, the others are results I don't want

Original.png
Fault.png
Fault2.png
User avatar
Dimitar - Tracker Supp
Site Admin
Posts: 1862
Joined: Mon Jan 15, 2018 9:01 am

Re: OCR Poor results

Post by Dimitar - Tracker Supp »

Hello Crookie,

Welcome to our Forum.

The OCR tool is not designed to recognize handwriting, but if you could give us a copy of the original document we will see what can be adjusted to get better results.

Regards.
Crookie
User
Posts: 3
Joined: Thu Jun 06, 2024 2:01 pm

Re: OCR Poor results

Post by Crookie »

This isn't just one document, this is just a sample I've been given.
We will be talking thousands, and it doesn't look like PDF-Xchange is up to it
User avatar
Tracker Supp-Stefan
Site Admin
Posts: 18265
Joined: Mon Jan 12, 2009 8:07 am
Location: London
Contact:

Re: OCR Poor results

Post by Tracker Supp-Stefan »

Hello Crookie,

Unfortunately the ABBYY Fine Reader engine that our Enhanced OCR uses is really focused on other types of text and handwritten recognition is not it's strength. Tesseract (the engine behind our standard OCR) - might be handling such text slightly better - so please do give that one a try as well. Unfortunately we can not really improve those OCR engines on our end, so if you have thousands of handwritten documents to OCR - we might not be able to fully help!

Kind regards,
Stefan
Post Reply