AWS Machine Learning Blog

Use two additional data labeling services for your Amazon SageMaker Ground Truth labeling jobs

We’re excited to announce the availability of two more data labeling services that you can use for your Amazon SageMaker Ground Truth labeling jobs:

  • Data Labeling Services by iMerit’s US-based workforce
  • Data Labeling Services by Startek, Inc.

These new listings on the AWS Marketplace supplement the existing iMerit India-based workforce listing to provide you a total of three options.

iMerit now provides access to their full-time, US-based staff of data labeling specialists. Their image labeling capabilities include classification, bounding boxes, image segmentation, key points, polygons, and polylines. Their text labeling capabilities include entity extraction and classification in both English and Spanish.

StarTek is a business process outsourcing company that offers data labeling services. StarTek is a publicly traded company (NYSE: SRT), and their workforces are spread across the Philippines, Honduras, India, Brazil, and Jamaica. Their image labeling capabilities include classification, bounding boxes, image segmentation, key points, polygons, and polylines. Their text labeling capabilities include entity extraction and classification in English.

We launched Amazon SageMaker Ground Truth at re:Invent 2018. It’s a service that helps you build highly accurate training datasets for machine learning. You can learn more from our launch blog. When you set up a Ground Truth labeling job, you can send labeling tasks to your own workers, Amazon Mechanical Turk public workers, or one of the vendors with listings on the AWS Marketplace.

You can assign data labeling tasks to one of the pre-approved vendors, who are vetted by Amazon for confidentiality, service guarantees, or special skills. The vendors are approved based on meeting specific requirements for data security, restricted access to physical facilities, and secure data transmission. We perform regular security audits to ensure the vendors continue to meet requirements.

Typically, finding and then contracting the right vendor is a time-consuming and tedious process. With Ground Truth, working with vendors is simple and involves just a few clicks through AWS Marketplace. All vendor-related charges appear directly on your AWS bill through the AWS Marketplace listing. The steps here show how easy it is to work with vendors to complete your Ground Truth labeling jobs.

Step 1: Navigate to the Vendor tab for Labeling Workforces

After signing into your AWS account, navigate to the Amazon SageMaker console. On the left-hand side navigation panel, select Labeling workforces. Then, choose Vendor on the right pane.

Step 2: Subscribe to the labeling services of a vendor through AWS Marketplace

After you choose Find data labeling services, you’re directed to the AWS Marketplace.

From here, you can select any of the available vendors to learn more about their company, labeling services, pricing, and much more. After you have selected the vendor that meets your needs, choose Continue to Subscribe on the listing page and complete the subscription process. Now you can use this vendor for a Ground Truth labeling job. You can be subscribed to any number of vendors at any time.

Step 3: Select a vendor when setting up your labeling job

When you create a labeling job, you see a list of all your subscribed vendors in the Subscribed data labeling services dropdown list. Choose one to kick off your labeling job. You have the flexibility to use different data labeling services for any of your labeling jobs.

Now Available

If you want to learn more about each of the data labeling services, visit the AWS Marketplace listings page. Now it’s your turn to work with them, and let us know what you think.


About the Author

Vikram Madan is the Product Manager for Amazon SageMaker Ground Truth. He focusing on delivering products that make it easier to build machine learning solutions. In his spare time, he enjoys running long distances and watching documentaries.