How to know if images are being labelled automatically?

SageMaker Ground Truth provides a few ways to check the status of your automatic and manual labelling tasks. Here are some methods:

• AWS Management Console: You can check the status of your labelling job from the Amazon SageMaker console. Navigate to the SageMaker Ground Truth console and then select the "Labelling Jobs" section. Each labelling job should show its status, including whether it's complete, in progress, or failed. You should be able to find out whether automatic labelling is enabled for your job by looking at the details of the job.

• AWS CLI: If you are comfortable using the command line, you can use the AWS CLI (Command Line Interface) to describe your labelling job and check its status. The describe-labeling-job command will give you detailed information about the labelling job, including the number of images that were labelled manually and the number of images that were labelled automatically. Here's an example of how to use this command:

aws sagemaker describe-labeling-job --labeling-job-name 'your-job-name'

Replace 'your-job-name' with the name of your labelling job. The output will be in JSON format, and you can look for the 'LabelCounters' key to find out the number of labelled images. Please note that you may need appropriate permissions to access these resources. If you don't have the necessary permissions, you might need to contact your AWS administrator.

Please also note that automatic labelling only happens once a sufficient number of manually labelled examples are available, and the quality of the automatic labelling is high enough. This is why a random sample of images is first sent to human workers for manual labelling. The exact number of manually labelled examples required depends on the complexity of the labelling task and the quality of the automatic labelling.

As I know, AWS doesn't provide a specific number or percentage of images that you are supposed to label manually when automatic labelling is enabled. This might vary depending on the specific circumstances of the labelling job.