Run layout doesnt recognize text

Question

Run layout doesnt recognize text

Louis-Philippe Redaud 65

When running layout, it doesn't recognize the 2 values below:

User's image

How to fix this? I tried to resave the file as pdf.
Any workaround?

thanks

navba-MSFT 27,605 Reputation points Microsoft Employee Moderator

2024-12-10T07:18:01.2266667+00:00
@louis-philippe redaud Thanks for Reaching the Microsoft Q&A Forum.

.

Please let me know which layout and its api-version are you using ?

Did you try checking from the Document Studio ?

Also please share the document which you are testing with.

Which region is your Document Intelligence resource deployed to ?

.

Awaiting your reply.
Louis-Philippe Redaud 65 Reputation points

2024-12-10T13:47:30.8266667+00:00

I'm using 2024-07-31 api version from document intelligence. I'm making my own custom extraction model. But I havethe same result when using the 'read' one.

DI ressource is in Easst US.

testtoromont.pdf
Louis-Philippe Redaud 65 Reputation points

2024-12-10T13:49:21.6066667+00:00

deleted
navba-MSFT 27,605 Reputation points Microsoft Employee Moderator

2024-12-13T10:07:24.15+00:00

@louis-philippe redaud I tried the same pdf at my end with the pre-built Read model in East US region and it worked just fine:

. Could you please test again with an image of the same pdf with better quality and check if that helps ?

.

Awaiting your reply.
Louis-Philippe Redaud 65 Reputation points

2024-12-13T17:57:43.73+00:00

On you screen shot the '1' on the 1st line is not detected. I

I tried to use a different API version, and it looks like the result are differents. With the API version 2024-02-29 I have the following:1st line is read correctly, but the '2' on the 3rd line is not read.

I can't improve the quality as that document is issue by the supplier and the pdf is clearly readable.
navba-MSFT 27,605 Reputation points Microsoft Employee Moderator

2024-12-18T07:07:21.9466667+00:00

@louis-philippe redaud Thanks for getting back. I am looking into this and I will get back to you once I have an update on this.
navba-MSFT 27,605 Reputation points Microsoft Employee Moderator

2024-12-26T04:03:10.4933333+00:00

@louis-philippe redaud Just following up to check if you had a chance to look at my above analysis. Awaiting your reply.

Answer accepted by question author

0 additional answers

Your answer

navba-MSFT 27,605 Reputation points Microsoft Employee Moderator

2024-12-10T07:18:01.2266667+00:00

@louis-philippe redaud Thanks for Reaching the Microsoft Q&A Forum.

.

Please let me know which layout and its api-version are you using ?

Did you try checking from the Document Studio ?

Also please share the document which you are testing with.

Which region is your Document Intelligence resource deployed to ?

.

Awaiting your reply.
Louis-Philippe Redaud 65 Reputation points

2024-12-10T13:47:30.8266667+00:00

I'm using 2024-07-31 api version from document intelligence. I'm making my own custom extraction model. But I havethe same result when using the 'read' one.

DI ressource is in Easst US.

testtoromont.pdf
Louis-Philippe Redaud 65 Reputation points

2024-12-10T13:49:21.6066667+00:00

deleted
navba-MSFT 27,605 Reputation points Microsoft Employee Moderator

2024-12-13T10:07:24.15+00:00

@louis-philippe redaud I tried the same pdf at my end with the pre-built Read model in East US region and it worked just fine:

. Could you please test again with an image of the same pdf with better quality and check if that helps ?

.

Awaiting your reply.
Louis-Philippe Redaud 65 Reputation points

2024-12-13T17:57:43.73+00:00

On you screen shot the '1' on the 1st line is not detected. I

I tried to use a different API version, and it looks like the result are differents. With the API version 2024-02-29 I have the following:1st line is read correctly, but the '2' on the 3rd line is not read.

I can't improve the quality as that document is issue by the supplier and the pdf is clearly readable.
navba-MSFT 27,605 Reputation points Microsoft Employee Moderator

2024-12-18T07:07:21.9466667+00:00

@louis-philippe redaud Thanks for getting back. I am looking into this and I will get back to you once I have an update on this.
navba-MSFT 27,605 Reputation points Microsoft Employee Moderator

2024-12-26T04:03:10.4933333+00:00

@louis-philippe redaud Just following up to check if you had a chance to look at my above analysis. Awaiting your reply.

Answer 1

navba-MSFT 27,605 Microsoft Employee Moderator

@louis-philippe redaud Apologies for the late reply. I had reached out to the Product Owners and got a reply from them.

.

Cause:

The '2' on the 3rd line is not read and it's likely related to dots in the background. This is recognized as 1: then rejected.

User's image

.

Please note that this is not a by-design behavior. Actually it's a quality issue that we need to improve. We are working on some improvements. However, the improvements will take few months as we need to working on model trainings.

.

Workaround:

As for workaround, if you have some way to remove those dots in the background of that row, that will help on the specific case.

.

Hope this answers.

Louis-Philippe Redaud 65 Reputation points

2025-01-14T16:22:54.17+00:00

Hi, thanks for the reply. I will check from time to time if a new API version is available to upgrade the project.

Share via

Run layout doesnt recognize text

0 additional answers

Your answer