Amazon announced that SageMaker Ground Truth, its service to help build highly accurate training datasets for machine learning, now offers simplified labeling workflows and, support for additional labeling vendors, and is available in the Asia Pacific (Sydney) AWS region, which brings the total to six supported AWS regions in the Americas, Europe, and Asia.
SageMaker Ground Truth launched at AWS re:Invent 2018, and offers a set of tools designed to be used with Amazon’s SageMaker service According to the blog post about the announcement from AWS Technical Evangelist Julien Simon, SageMaker Ground Truth can lower customers' labeling costs by up to 70% using automatic labeling. Furthermore, with SageMaker Ground Truth customers can build datasets for image classification, object detection, semantic segmentation and others.
Source: https://aws.amazon.com/sagemaker/groundtruth/
Since the release of SageMaker Ground Truth, Amazon has received feedback from various customers such as T-Mobile, Pinterest, Change Healthcare, GumGum, and others. With the input from these customers, Amazon further updates the SageMaker Ground Truth with features such as:
- Job chaining - to help developers save time by allowing them to run subsequent machine learning labeling jobs by using the output of previous jobs
- Job tracking - to enable developers to track the progress of labeling jobs in real time
- Long-lived Jobs - enabling human experts to be used as labellers and update models periodically
- Dynamic Custom Workflows - allowing additional context to be added to the source data, meaning developers can add input from previous labeling jobs
With these features, T-Mobile can have its data scientists focus more on other tasks. Cody Sanford, executive vice president and chief information officer at T-Mobile, said in a WSJ article:
We were spending copious amounts of our data scientists’ time labeling thousands upon thousands of messages. Ground Truth has been able to make that super-efficient, and now we don’t need our skilled data scientists hand-labeling data anymore.
Additional to these features, Amazon is working with new service providers, including Vivetic and SmartOne, to add support for data labeling in French, German and Spanish.
Amazon SageMaker Ground Truth is now available in six AWS regions, including Virginia, Ohio, Ireland, Tokyo and Sydney. Furthermore, pricing details of Amazon SageMaker Ground Truth are available on the pricing page.
