Label Verification - Amazon SageMaker

Label Verification

Building a highly accurate training dataset for your machine learning (ML) algorithm is an iterative process. Typically, you review and continuously adjust your labels until you are satisfied that they accurately represent the ground truth, or what is directly observable in the real world.

You can use an Amazon SageMaker Ground Truth label verification task to direct workers to review a dataset's labels and improve label accuracy. Workers can indicate if the existing labels are correct or rate label quality. They can also add comments to explain their reasoning. Amazon SageMaker Ground Truth supports label verification for Bounding Box and Image Semantic Segmentation labels.

You create a label verification labeling job using the Ground Truth section of the Amazon SageMaker console or the CreateLabelingJob operation. Ground Truth provides a worker console similar to the following for labeling tasks. When you create the labeling job with the console, you can modify the images and content that are shown. If you create a labeling job using the API, you must supply a custom-built template. To learn how to create a custom template, see Creating Custom Labeling Workflows. To see examples of custom templates that can be used for label verification labeling job types, see this Github Repository.

You can create a label verification labeling job using the Amazon SageMaker console or API. To learn how to start a label verification job on the console, see Getting started. To use the API, see CreateLabelingJob.