To run repeat image checks, the service requires historic data to compare against. Hence, after an instance is provisioned, it should be indexed to create a historic database.

Note: Indexing an instance is an optional step. However, the service will not produce any matches on an un-indexed instance since it does not have any a-priori information to compare against.

How a Document is Indexed

A submitted image is converted into a vector by utilizing embedders. This vector is then stored in a database to create the historic dataset. This embedding is non-reversible. The vector cannot be converted back into the original image.

The response of Index Record API is stateless. Since the values are hashed when stored in the historic database, this response can never be regenerated.

Refer to the Index Images API for further details.