Posted On: Mar 14, 2019

Amazon SageMaker Ground Truth now supports multiple categories for the bounding box workflow and three new user interface (UI) templates for setting up custom labeling workflows.

Amazon SageMaker Ground Truth helps you build highly accurate training datasets quickly. SageMaker Ground Truth offers easy access to public and private human labelers and provides them with built-in workflows and interfaces for labeling tasks. Additionally, SageMaker Ground Truth can lower your labeling costs by up to 70% using automatic labeling, which works by training Ground Truth from data labeled by humans, so that the service learns over time to label data independently.

The bounding box workflow now allows you to label multiple categories within an image simultaneously. For example, you can now ask labelers to draw bounding boxes around pedestrians and cars in the same image. As a result, you can now create more efficient labeling jobs for sophisticated computer vision models.

Also, SageMaker Ground Truth now supports three new UI templates for your custom workflows: draw polygons around areas of interest in an image, mark key points in an image, and segment pixels of each unique instance of an object in an image. With these new UI templates, you can now choose from 15 different templates to get you started quickly, in addition to the option of creating your own template with custom workflows.

You can get started with Amazon SageMaker Ground Truth here, and refer the developer guide for additional information.