Amazon SageMaker Ground Truth announces new features to simplify workflows, new data labeling vendors, and expansion in the Asia Pacific region
Amazon SageMaker Ground Truth offers easy access to public and private human labelers and provides them with built-in workflows and interfaces for common labeling tasks to provide significant speed and cost benefits to labeling data for machine learning. SageMaker Ground Truth now offers simplified labeling workflows, support for additional labeling vendors, and has been extended to a sixth AWS region, making it even easier to build highly accurate training datasets. Successful machine learning models are built on the shoulders of large volumes of high-quality training data.
Amazon SageMaker Ground Truth offers easy access to public and private human labelers and provides them with built-in workflows and interfaces for common labeling tasks to provide significant speed and cost benefits to labeling data for machine learning. SageMaker Ground Truth now offers simplified labeling workflows, support for additional labeling vendors, and has been extended to a sixth AWS region, making it even easier to build highly accurate training datasets. Successful machine learning models are built on the shoulders of large volumes of high-quality training data.
Within workflows, you can now chain together labeling jobs with a single click. Chaining labeling jobs enables you to use the output from a previous job or from a previous machine learning model as the input for a subsequent job. As a result, you can create more accurate training datasets. For example, if a labeling job identified humans in an image, the subsequent job could draw bounding boxes around the humans, making the labeling more accurate.
Custom workflows created in Amazon SageMaker Ground Truth can also now inject the output from previous labeling jobs or other relevant content into your custom labeling workflows. As a result, you can provide labelers with additional context to help them complete labeling jobs faster and with greater accuracy. For example, you can display the outside temperature for each image in a weather classification task to help the labeler classify the image.
For all workflows, you can now track the progress of your labeling jobs in the console in near real-time. Also, every batch of a labeling job can now run for up to 10 days.
In addition to simplified workflows, we are also announcing support for two additional labeling vendors in the AWS Marketplace, Vivetic and SmartOne, bringing the number of approved vendors to four. With these additional vendors, data labeling is now supported in French, German, and Spanish.
Lastly, Amazon SageMaker Ground Truth is now available in the Asia Pacific (Sydney) AWS region, bringing the total to 6 supported AWS regions in the Americas, Europe, and Asia.
You can refer to the Amazon SageMaker Ground Truth documentation for details, and also learn more in the blog post.