ai-computer-vision
2021.10
false
- Release notes
- Overview
- Setup and configuration
- Installation details
- Deploying the package
- Data storage
Deploying the package
AI Computer Vision User Guide
Last updated Jun 27, 2024
Deploying the package
This page describes how to deploy an out-of-the-box Computer Vision ML package.
To create a Computer Vision ML package, select the AI Center service from the left navigation bar, create a new project or access an existing one, and then select ML Packages.
- Select Upload zip file.
- Under Package name, enter a package name.
- Under Upload package, upload the Computer Vision ML package as a
.zip
file.To do so, you need to download thecv-ondemand.tar.gz
bundle. For further details, go here. - Under Input type, select Json from the drop-down menu.
- Under Language, select the corresponding Python language from the drop-down menu.
- Enable the Recommend GPU toggle button.
- Select Create.
In the same project that you created the ML package, you need to:
- Go to ML Skills.
- Select Create new.
- Under Name, enter a name for the ML skill.
- Under Choose Package, select the ML package you just created from the drop-down menu.
- Under Choose package major version, select a major version from the drop-down menu.
- Under Choose package minor version, select a minor version from the drop-down menu.
- Enable the Enable GPU toggle button.
- Enable the Make ML Skill Public toggle button.
- Select Create.
- Once the ML skill is available, this can take up to 30 minutes, select the ML skill and copy the Url.
On activity level in your UiPath Studio project, you need to change:
- the URL property value of the CV Screen Scope activity, and
- the Endpoint property value of the UiPath Screen OCR activity
to the URL of the ML skill.
The default URLs can also be changed from the Project Settings page.
To change the endpoint for CV Screen Scope, go to UiPath Studio > your
project > Project Settings > Computer Vision > CV Screen
Scope > Server > Run value/Debug value > change
https://cv.uipath.com
to the URL of the ML skill.