Rekognition / Client / start_project_version



Starts the running of the version of a model. Starting a model takes a while to complete. To check the current state of the model, use DescribeProjectVersions.

Once the model is running, you can detect custom labels in new images by calling DetectCustomLabels.


You are charged for the amount of time that the model is running. To stop a running model, call StopProjectVersion.

For more information, see Running a trained Amazon Rekognition Custom Labels model in the Amazon Rekognition Custom Labels Guide.

This operation requires permissions to perform the rekognition:StartProjectVersion action.

See also: AWS API Documentation

Request Syntax

response = client.start_project_version(
  • ProjectVersionArn (string) –


    The Amazon Resource Name(ARN) of the model version that you want to start.

  • MinInferenceUnits (integer) –


    The minimum number of inference units to use. A single inference unit represents 1 hour of processing.

    For information about the number of transactions per second (TPS) that an inference unit can support, see Running a trained Amazon Rekognition Custom Labels model in the Amazon Rekognition Custom Labels Guide.

    Use a higher number to increase the TPS throughput of your model. You are charged for the number of inference units that you use.

  • MaxInferenceUnits (integer) – The maximum number of inference units to use for auto-scaling the model. If you don’t specify a value, Amazon Rekognition Custom Labels doesn’t auto-scale the model.

Return type:



Response Syntax


Response Structure

  • (dict) –

    • Status (string) –

      The current running status of the model.


  • Rekognition.Client.exceptions.ResourceNotFoundException

  • Rekognition.Client.exceptions.ResourceInUseException

  • Rekognition.Client.exceptions.LimitExceededException

  • Rekognition.Client.exceptions.InvalidParameterException

  • Rekognition.Client.exceptions.AccessDeniedException

  • Rekognition.Client.exceptions.InternalServerError

  • Rekognition.Client.exceptions.ThrottlingException

  • Rekognition.Client.exceptions.ProvisionedThroughputExceededException