Azure Doc Intel Run Layout bug

Hi EndCap

I am able to replicate the above observation on these two prebuilt models- layout and General documents

Issue behind not detecting the pointed-out region is it is digital checked, not checked and scanned.

We might need to take a printout and do high quality scan of the pdf or do a color thresholding of concerned pdf file.

Trying to come with a work around with above observation.

Please confirm details on prebuilt model name, version (v3.1 or 4.0) and region too.

Thank you

@Manas Mohanty This is a custom data extraction not a prebuilt model. The custom AzureModelID is customColoradoDisclosures. We're using the API 2024-11-30 (4.0 General Availability).

Hi EndCap

Thank you for clarifying. But issue is the quality of pdf files.

Could you provide a higher quality pdf files in private chat.

Thank you.

Hi EndCap,

Thank you for contacting Q&A Forum. I would like to provide my findings and proposed solution:

In your screenshot, the checkbox you're referring to isn't being highlighted in yellow, which means the model isn't detecting it as a selectable field. This limitation is known behavior with prebuilt models. To improve detection of all checkboxes you need, I recommend switching to a Custom model using Custom Labels. This allows you to:

• Manually tag the checkbox fields during training.
• Train the model to recognize similar checkboxes across other documents, even if they differ in style or placement.

See Generating labeled datasets for more information.

This approach gives you much more control over what is recognized, especially for documents with form-like layouts and checkboxes.

Kindly let me know if this work for you and please let me know if you have any further questions.

If I have answered your question, please accept this answer as a token of appreciation and don't forget to give a thumbs up for "Was it helpful"!

Best regards,

@Leo Tran I'm using custom extraction models and not pre-built models. We have manually tagged each data point we need including this checkbox. On my forms I have all other checkboxes reading in code except this one. Come to find out I notice Document Intelligence is not analyzing this field at all. There are also other random checkboxes that got missed except I don't care about those ones for now. The custom model ID is customColoradoDisclosures.