Release notes

To learn what has changed, check out the release notes.

v23.4.0

Release date: 26 April 2023

Improvements

Improved OCR detection.

Faster speed on text-heavy images.

The maximum number of words to be extracted per request has been increased from 800 to 2,000 to match the cloud and on-premises models.

Known issues

When the limit for the maximum number of words to be extracted per request is reached, an error message is thrown and the OCR boxes in the output are a random subset of the entire set of OCR boxes (words) on the screen, so we can’t guarantee this exact subset is detected at runtime.

As a workaround, you can either indicate a smaller region of the screen or you can disable OCR processing.

To disable OCR processing, if OCR boxes are not useful in the automation project, go to Project Settings > Computer Vision > CV Methods > deselect the OCR checkbox from the drop-down menu.

OCR processing can also be disabled at activity level if you go to the properties panel of the CV Screen Scope activity > Input > CvMethod > deselect the OCR checkbox from the drop-down menu.

v22.7.2

Release date: 25 July 2022

Bug fixes

• Fixed an issue which caused UiPath Screen OCR to crash when used as a standalone activity with the Computer Vision Local Server, or inside a CV Screen Scope that used either Computer Vision Cloud or Local Server.

v22.7.1

Release date: 12 July 2022

Improvements

The Computer Vision Local Server has now been upgraded and is faster, better, and shares all of its features with the Cloud version. Also, this means that older CPU architectures that do not support AVX2 are no longer supported and an error is thrown if trying to use the server on these machines.

v21.10.1

Release date: 21 January 2022

Bug fixes

• Fixed an issue which caused the Computer Vision Local Server to label some classes of UI elements (minimize buttons, close buttons, icons, check boxes, and input boxes) incorrectly.

v21.10.0

Release date: 9 November 2021

Improvements

This package now supports .NET 5 Windows projects. For more information, see About Automation Projects.

Breaking changes

• The Computer Vision Local Server does not correctly label UI Elements, causing automation projects created with earlier versions and ran on this version to potentially fail. This means that automations created with this versions do not have neither backwards or forward compatibility.

Important: Erratum 19 January 2022: Added the "Local Server does not correctly label UI Elements" breaking change.

v21.6.0

Release date: 29 June 2021

Improvements

The detection speed for UI elements is now improved between twofold and threefold, depending on the size of the screen you are using. Also, the new Machine Learning model now has increased accuracy when detecting UI elements.

You can now perform both Computer Vision and OCR detection in a single server call when using the CV Screen Scope activity with the OCR Screen Engine with both UseLocalServer properties set to True. The joint Computer Vision and OCR detection time is comparable to the Computer Vision detection time of the previous version.

This version of the Computer Vision Local Server does not support .NET 5. This means that when starting a new .NET 5 compatible Studio process, the Computer Vision Local Server package cannot be found on any feed, as any incompatible packages are filtered out by the Package Manager.

v20.10.0

Release date: 14th October 2020

What's new

The first iteration of the Computer Vision Local Server is now here, enabling you to harness the power of the Computer Vision activities without needing to connect to a third-party server.