AWS updates SageMaker Ground Truth to simplify ML training

Amazon Web Services (AWS) has updated Amazon SageMaker Ground Truth to make machine learning tasks a little easier. That’s what Silicon Angle reports.

SageMaker Ground Truth is a set of tools designed to work with Amazon SageMaker, which were unveiled last year. SageMaker is a service used by developers to build machine learning models for predictive and analytical applications.

The Ground Truth tools should make it easier to label datasets used to train those machine learning models. It allows developers to create more accurate datasets to train models for tasks such as image classification, object detection and semantic segmentation. The service essentially offers workflows and interfaces with which labelling tasks can be automated.

Updates

Now SageMaker Ground Truth gets a series of new features. Many of these have been developed in response to demand from existing users, says AWS Technical Evangelist Julien Simon. This includes a new job chaining feature that allows developers to run consecutive machine learning labeling tasks based on the output of previous tasks. This should allow them to save time.

There is also a new task tracking feature that allows developers to track the progress of their labeling tasks in real time. A feature called long-lived jobs allows human experts to be used as labelers and models to be updated on a periodic basis.

Finally, there is a new dynamic custom workflow feature. This makes it possible to add extra context to the source data. Developers can thus use the function to add input from previous label tasks.

New providers

Amazon not only announced updates, but also said it is working with new service providers to add more support for labelling data in French, German and Spanish. These include Vivetic and SmartOne.

Amazon SageMaker Ground Truth is also being made available in more areas, including Virginia, Ohio, Ireland, Tokyo and Sydney.