Amazon Rekognition improves the accuracy of image moderation

Posted on: Oct 8, 2018

Amazon Rekognition is a deep learning-based image and video analysis service that can identify objects, people, text, scenes, and activities, as well as detect unsafe content. Amazon Rekognition now comes with an improved image moderation model that reduces false positive rates by 40% on average without any reduction in detection rates for truly unsafe content. Lower false positive rates imply lower volumes of flagged images to be reviewed further, leading to higher efficiency of human moderators and more cost savings.

With Amazon Rekognition's image moderation feature, you can automatically review photo libraries of any scale to detect unsafe content. Human moderators only need to review a much smaller set of images that have been flagged as ‘Explicit Nudity’ or ‘Suggestive’ content by Amazon Rekognition. This allows your workers to focus on more valuable activities and still achieve full moderation coverage at a fraction of the cost. Moreover, Amazon Rekognition provides a hierarchical set of moderation labels that can be used to create business rules to handle different geographic and demographic requirements. For more details on supported labels, please see this page.

“Amazon Rekognition makes it easy for us to detect and take action on user-generated photos that need moderation. The DetectModerationLabels API provides the intelligence for us to make automated decisions on how to best moderate our content with high confidence. With Amazon Rekognition, we were able to reduce the need for human involvement by 97%, which enabled our employees to focus on more strategic business tasks and lowered our moderation costs by 72%. Amazon Rekognition also helped us decrease the time required for photo approvals from hours to minutes while maintaining the same accuracy.”

- Maksim Gutman, Engineering Manager, Coffee Meets Bagel.

Improved image moderation is now available in all AWS Regions supported by Amazon Rekognition at no additional cost. Improved video moderation will be available in the next 6-8 weeks. To get started, you can try the latest version in the Rekognition Console.