How to extract data from a video using Azure AI Video Indexer?

Athento allows the extraction of information from a video using Azure AI Video Indexer.

To perform this extraction, the first step is to configure the connection with Azure AI Video Indexer. To do this, the connection object must be created from the administration in the ‘Home > Athentose_Azure > Azure video indexer configs’ section.

The attributes to configure are:

• Subscription_key: Subscription key. This is obtained from the Azure AI Video Indexer platform itself.
• Account_id: Account ID. This is obtained from the Azure AI Video Indexer platform.
• Location: Azure AI Video Indexer account location.
• Team: Not required. Indicates the team for which the connection is configured.

Video sending operation

In the marketplace of the space automations, the ‘Send document to azure video indexer’ operation must be installed, which has a parameter to be configured:

• Azure video indexer config UUID: UUID of the Azure video indexer config object configured in the previous section.

Once this operation has been configured, it must be launched on a document to start processing the video.

**NOTE: The document must be a video for the processing to take place, otherwise the operation will do nothing. The allowed video extensions are ‘mp4’, ‘mov’ and ‘ogg’.

What happens once Azure has finished processing the video?

When Azure finishes processing the video, a callback will be received from Azure to the Athento instance. This call will execute the operations with the category ‘azure_video_indexer_callback’ on the document from which the operation was launched. To define the category of an operation, you must go to the advanced administration of the operation and set the corresponding category.

Callback operations

Athento currently has a callback operation ‘op_azure_video_indexer_callback_save_named_entities’ which allows to place the data obtained from the video on a JSON type metadata.

Parameters:

• JSON Field: JSON field on which the data will be placed.

The data to be placed as metadata value are:

• brands: Marks detected in the video.
• namedLocations: Known locations detected in the video.
• namedPeople: Known people detected in the video.
• topics: Topics that the video deals with.

It is important to put the category ‘azure_video_indexer_callback’ to the operation as mentioned above.