How do I configure the SageMaker Ground Truth verification and adjustment jobs?

Question

I learnt from the Amazon SageMaker Ground Truth documentation that there is the possibility to run [label verification and adjustment jobs][1]. I have the following questions:

1. What's the pricing for these options? Are they similar to the pricing for the labeling jobs?
2. With multiple individuals in the workforce, which one of the following is true?
The labeled dataset is split across multiple individuals.
A complete dataset is considered per individual doing the verification.
3. Can I use Ground Truth to select a random subset of the labeled dataset for verification?

[1]: https://docs.aws.amazon.com/sagemaker/latest/dg/sms-verification-data.html#sms-data-verify-start-api

Accepted Answer

The following are the responses for your questions:

1. Adjustment and verification jobs are just types of labeling jobs. You can configure an adjustment or verification job through the CreateLabelingJob API or through the [Create labeling job][1] option in the Amazon SageMaker console. The same pricing per dataset object (i.e., per image or text document) applies.
2. You can configure these jobs. When you set up a labeling job, you provide the work team ARN. You can configure this work team however you see fit. Typically, customers who want to do the adjustment or verification include only the individuals who are highly trained reviewers or managers. After you create the labeling job, the verification or adjustment tasks are served to the members of the work team.
3.  When you create a labeling job in the SageMaker console, use the **Additional configuration** option to configure Ground Truth to select a randomized subset.

[1]: https://docs.aws.amazon.com/sagemaker/latest/dg/sms-create-labeling-job-console.html

How do I configure the SageMaker Ground Truth verification and adjustment jobs?

相關內容