View a markdown version of this page

Use input and output data - Amazon SageMaker AI
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Use input and output data

Note

After careful consideration, we have made the decision to close new customer access to Amazon Ground Truth, effective 7/30/26. Existing customers can continue to use the service as normal. Amazon continues to invest in security and availability improvements for Ground Truth, but we do not plan to introduce new features.

The input data that you provide to Amazon SageMaker Ground Truth is sent to your workers for labeling. You choose the data to send to your workers by creating a single manifest file that defines all of the data that requires labeling or by sending input data objects to an ongoing, streaming labeling job to be labeled in real time.

The output data is the result of your labeling job. The output data file, or augmented manifest file, contains label data for each object you send to the labeling job and metadata about the label assigned to data objects.

When you use image classification (single and multi-label), text classification (single and multi-label), object detection, and semantic segmentation built in task types to create a labeling job, you can use the resulting augmented manifest file to launch a SageMaker training job. For a demonstration of how to use an augmented manifest to train an object detection machine learning model with Amazon SageMaker AI, see object_detection_augmented_manifest_training.ipynb. For more information, see Augmented Manifest Files for Training Jobs.