Integration with AI

Owned by Daniel Parshin
Last updated: Dec 14, 2023 by Jared Taylor
2 min read

VXG aggregates and pre-processes video for AI and provides the video data in many formats including:

  1. Images on demand and at any sample rate

  2. Clips on demand and of any length

  3. Streaming live and recorded video

VXG integrates with AI engines using a so-called group access token - a string variable that contains all information about a collection of cameras that are assigned to work with this AI.  

An AI engine uses the group access token to enumerate cameras and call VXG functions to retrieve data required by the AI engine. It could be images, clips, live or recorded streaming video. Using a group access token you can get access tokens for each individual camera and call functions for each camera individually. At the same time, some functions support operations for a group of cameras and in this case, a group access token is used to perform an operation with a collection of cameras.

Integration steps

  1. Configure which cameras should be processed, this step includes

    1. Creating a group access token.

    2. Adding cameras to the group access token.

  2. Passing the group access token to the AI service. This is just passing a long string from the Cloud VMS to the AI service using any available method (from API to just hardcoding the string).

  3. Retrieving data from channels for AI processing, this step includes

    1. Retrieving individual channel access tokens from the group access token.

    2. Retrieving live or recorded video, clips or images for each channel.

  4. (Optional) Recording metadata generated by AI to the Cloud VMS storage.

Access data from an AI engine

AI engine needs to know a group access token. Using the group access token the AI engine can access data from any camera in the collection that is associated with this particular AI. See for samples of how to work with group access tokens.

See  for samples of how to generate images from live video, at any rate, batches of images from video storage, and how to create clips.

For samples of how to work with metadata, please refer to

The full API is published in the REST API section.

Sample code

There are many sample codes published on

The following sample code - ai_integration.py will show usage of the main required functions:

  • list_channels_for_lkey - shows all channels for the provided LKey.

  • create_group_channel - creates a group channel and returns its ID and group access token

  • add_channel_to_group - adds channels to a group channel.

  • get_channels_from_group - get channels from a group channel

  • get_live_urls_for_channels_in_group - retrieve live URLs for channels in a group channel

  • record_metadata_to_channel - records metadata to a channel

 

 

Related documents: