Celebrity spotting: Amazon improves facial recognition for famous people

Amazon promises its Rekognition machine-learning tech will save hours of humans tagging photos and videos of celebrities.
Written by Liam Tung, Contributing Writer

Rekognition for video facial recognition analysis has been available since 2017.

Image: Amazon

Facial recognition AI used by police on the public remains a contentious issue, but Amazon is pressing ahead with its Rekognition technology for identifying celebrities in images and video.

In May, Amazon extended its one-year ban from June 2020 on US law enforcement using Rekognition facial recognition until sufficient regulations are introduced. IBM exited the business altogether, while Microsoft has paused sales to police departments until federal laws are passed.

But one application where Amazon Rekognition is still valid is celebrities, who are already in the public eye. The service caters to media companies, and AWS says that Rekognition now "automatically recognizes tens of thousands of well-known personalities in images and videos using machine learning (ML)."

SEE: The Privacy Paradox: How can businesses use personal data while also protecting user privacy?

AWS has updated its models to provide more accurate detections and greater coverage of global celebrities. It also added three new metadata attributes to celebrity profiles -- gender, expression, and smile -- to help users search for specific moods of celebrities in stills and video. So, instead of a human spending hours tagging photos, AWS will tag them automatically across large media asset repositories.  

Similar facial recognition services include Microsoft's Azure Face and Google's Cloud Vision API, though it currently only offers a restricted and beta version of its facial celebrity recognition technology. Google's service is also aimed at entertainment and media companies.

Rekognition for video facial recognition analysis has been available since 2017. Amazon Rekognition Video uses deep learning to analyse video for recognisable objects, scenes, celebrities, text, activities, and inappropriate content from customers' videos stored in Amazon S3.

The celebrity recognition feature promises to save media firms the cost of humans tagging media, while also making it easier to search images.

"With the rapid proliferation in the volume of image and video content available on video on demand (VOD), streaming, and social media platforms, media companies are struggling to organize, search, and fully use their catalogs at scale," AWS says in a blog post.  

"Similarly, news channels often need to rapidly locate images and videos of a particular celebrity in response to current events, but insufficient metadata makes it tedious to search their libraries for the right clips. Sports broadcasters also find it hard to quickly find the right footage from games and interviews to create highlights, shorts, and special programming." 

Supported AWS regions for the celebrity recognition feature include the US, Asia Pacific, Japan, and Europe.    

Editorial standards