Amazon Rekognition Gets Better at Detecting Faces
Amazon Web Services (AWS) recently improved its Amazon Rekognition AI solution, expanding its ability to identify faces in crowded photos, as well as adding a new text-identification feature.
Released a year ago at the 2016 AWS re:Invent conference, Rekognition is an image-recognition service that lets users search for and index real-world faces, objects and scenes in photos. It also has the ability to detect attributes like gender, expression and distinguishing facial characteristics.
Last week, AWS announced an upgrade to the solution's facial-recognition capabilities. Rekognition can now search for specific faces in real time against catalogs of millions of photos. "This represents a 5-10X reduction in search latency, while simultaneously allowing for collections that can store 10-20X more faces than before," AWS said in its announcement.
The real-time search capability can be particularly helpful for law enforcement agencies, which are constantly looking for ways to reduce the time it takes to identify suspects, AWS noted.
Rekognition also now has the ability to identify a higher number of faces in crowded or busy photos -- up to 100 faces, compared to the previous maximum of 15. Businesses like department stores can use this capability to gain insights into customer sentiment or demographics.
AWS has also added the ability for Rekognition to detect text (such as license plates or street signs) that are captured in real-world images (ranging from photos to stills from TV reports to social media posts).
This new "Text in Image" feature also enables users to catalog and search for images that display the same text. It "supports text in most Latin scripts and numbers embedded in a large variety of layouts, fonts, and styles, and overlaid on background objects at various orientation as banners and posters," AWS said.
Gladys Rama (@GladysRama3) is the editor of Redmondmag.com, RCPmag.com and AWSInsider.net, and the editorial director of Converge360.