Document Understanding - 21.6 Community Preview

Greetings, UiPath Community! :dancer:

Lately we have continued the series of improvements on the Validation Station (this time for the Activities used in the Action Centre), performance as well as other fixes here & there.

As a summary, we have worked on:

Showing only Classified Page Ranges in the Validation Station of the Action Center

With this release, we have enabled the “Create Document Validation Action” to send only the relevant page ranges to the Validation Station in the Cloud, so that Knowledge Workers using the Action Center can focus on only the document part of their interest, instead of seeing extra pages which they do not need to review or process!

This also results in increased performance, as only relevant document pages, dom and extraction results are sent to and from the Action Center, to the “Wait For Document Validation Action And Resume” Activity - where they will be picked up for further processing.

Note: The ShowOnlyRelevantPageRange option can only be configured on the “Create Document Validation Action”, having the “Wait For Document Validation Action And Resume” using the same value for it (as configured previously).

Performance Improvements - OCR as Local Server

We have worked on improving the general performance of the product. Among others in this sense, we have enabled RPA Developers to use locally installed packages of OCR engines for Documents and Screen Scraping, providing for the two corresponding OCR activities (Screen OCR and Document OCR) the option to “UseLocalServer”, not requiring an AI Center hosted model, but rather running on the robot’s local machine. The activities require the UiPath.ComputerVision.LocalServer and UiPath.DocumentUnderstanding.OCR.LocalServer respectively package installed to use the local server mode.

Exporting ML Extractor Training Data into a Dataset

In addition to the Output Folder, we are now providing the possibility of sending data from Validation Station directly to a Dataset in your AI Center Tenant.

When using the Dataset option, the activity will send the data to a folder called “fine-tune” within the selected Dataset, where the Data Manager expects it to be and from where it imports (and removes, after import - so do use both the AI Center Dataset and the local folder property if you want a backup for it).

Keeping up to date with a .net 5 upgrade

We have adapted the OCR.Activities and OCR.Contracts to support .net5 framework.

The usual - stabilization & bug fixing

We have worked on small improvements of the digitization results and boxes around them, as well as the UX of the Template Manager and the Validation Station.


I’m using the 21.6. Beta studio (Enterprise).

How can I update my activities that the UiPath Document OCR can run locally?
UiPath.OCR.Activities are currently v3.0.0

Other than that, nice features!

Hi @T0Bi

To run UiPath Document OCR locally, you need to:



I hope i can write here for this topic, in the data extraction scope, it is possibile, if a value is not extracted to put another value inside ( e.g. a value extracted from a db)? if yes how? Cause I’ve experimented only to modify a variable if this is extracted.

Thank you so much in andvance,


Thanks… I just figured out that I forgot to check the “Include prerelease” checkbox.

Hi Marica!

Currently, you can modify the values of the extractions by manipulating the Extraction Results programatically or by creating your own Extractor (see samples here: GitHub - UiPath/Document-Processing-Code-Samples: Code samples for document processing activities.)

Hope it helps,

1 Like

Thanks for the share!