Amazon Rekognition Gains Additional Features

Amazon Rekognition, the auto metadata extractor that launched in 2016, has recently received some much needed love.

The cloud-based software launched by AWS has been a significant tool for image and video analysis since its inception, being used by thousands of companies worldwide (including us). Here’s Vanja Josifovski, CTO at Pinterest, speaking about the technology (a fascinating watch)!

Unfamiliar with Amazon Rekognition? It essentially simplifies the process of image and video analysis interpretation to cloud applications using highly scalable, deep learning technology and machine learned principles. It factors in GitHub solution templates to deploy image and video analysis faster and more efficiently by developers and has been used for many verticals.

Here’s just a few:

• Video segment detection – Great for video archives that need to detect key areas in their thousands of hours of footage. Helping to remove black frames, unwanted credit sections, shot boundaries and so on.

• Celebrity/Individual recognition – Identify well known celebrities or targeted individuals by scanning catalogues libraries and interpreting faces, postures and poses.

• Text detection – Scan and read text within images and videos, even when skewed by movement and distortion.

• Content moderation – Making the internet a safer place to surf by flagging potentially inappropriate content and unwanted media from appearing on sites, apps and more. Can even be bolstered by the use of Amazon A2I to automate human intervention steps if something gets flagged.

So what are the new features?

Well as of a few days ago, AWS have completely updated the list of object bounding boxes and labels that users can apply to their own use cases and applications.

The label detection feature of Amazon Rekognition is a heavily used resource, providing truly rich and useful metadata for faster automation (and less of a technical battlefield to sift through 😅 ). The best example of how this increase in features will affect particular business types would be to look at a social media platform:

If a user was to search for ‘ocean‘ or ‘football‘, this increase in labels and bounding boxes would make it simpler to automate search results that have a higher % of satisfying the users needs.

There are however an unlimited number of other use cases, such as how a marketing department could use the data to analyze which visual elements in a creative led campaign led to the most engagement and conversion. Very exciting stuff!

Keen to learn more about how your business could use Amazon Rekognition?

We have experience in deploying Amazon Rekognition as well as other unifying services like Amazon Textract for our clients. Get in touch with us today to speak with a member of our team.