OCR Reading document in wrong perspective

Help AI Center
,
1

Hello Everyone,

I am having a problem reading a document in the labeling part of Document Understanding.

Screenshot_2
Screenshot_21036×373 151 KB

As you can see on the print screen, the OCR understands only the text on the bottom left of the document and ignores all the other texts.

What can I do to make the OCR read all the information in the document?

Thanks in advance.

2

Hi @Samuel_Simao ,
Can you let me know which OCR are you using?

I recently came across a similar issue and switched to UiPath Document OCR. They have internally designed to realign the image/pdf and digitize the document accordingly.

Let me know if this helps

Thanks,
Nishant

Thanks,
Nishant

3

Hey @Nishant_Banka1 ,

Thank you for your atention.

I am currently using UiPath Document OCR as well.

image
image926×475 76.8 KB

4

Hi @Samuel_Simao,
I have no clue then. I don’t think there is an option provided to rotate the image in the labeling session.

One way of getting through would be to delete this particular file, rotate it on your local machine and then import the rotated file back to your labeling session. This would work if there’s only a handful of such files, but if all your files are like this, then I don’t think this approach would be the best suited.

Thanks,
Nishant

5

I found a solution. For some reason that I don’t know, when I digitize the document with the UiPath Document OCR in the Studio, the document is read perfectly. So I use that and the train extractors scope to send the documents to the data manager.