3 min read

According to Mary Meeker’s 2016 Internet trends report, we are now sharing a staggering 3.25+ billion digital photos every day. In the era of smartphones, the challenge for organizations is to index and interpret this data. Amazon tried to solve this problem with its deep learning-powered Rekognition service which it unveiled at last year’s AWS re:invent conference.

By June this year, Amazon Rekognition had become a lot more smarter, recognizing celebrities across politics, sports, business, entertainment and media.

Now Rekognition has truly arrived – it can ‘recognize’ textual images and faces at real time! Amazon has infused three new features into the service: detection and recognition of text in images; real-time face recognition across tens of millions of faces; and detection of up to 100 faces in challenging crowded photos.

The new functionalities, Amazon claims, make Rekognition “10% more accurate” for face verification and identification.

Text in Image

Being able to detect text in images is, in fact, one of the most anticipated features that have got added into Rekognition. Customers have been pressing about recognizing text embedded in images, such as street signs and license plates captured by traffic cameras, news, and captions on TV screens, or stylized quotes overlaid on phone-captured family pictures.

Well the system can now recognize and extract textual content from images.

Interestingly, the Amazon Web Services announced that Text in Image is specifically built to work with real-world images rather than document images. “For example, in image sharing and social media applications, you can now enable visual search based on an index of images that contain the same keywords. In media and entertainment applications, you can catalogue videos based on relevant text on screen, such as ads, news, sport scores, and captions. Additionally, in security and safety applications, you can identify vehicles based on license plate numbers from images taken by street cameras,” AWS said in its official release.

The Text in Image feature 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.

Face Search and Detection

With Amazon Rekognition, customers can now perform real-time face searches against collections of millions of faces. “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.

The face search feature can truly prove to be a boon in security and safety applications – for timely and accurate crime prevention – where the suspects can be identified against a collection of millions of faces in near real-time.

On top of all that, Rekognition now allows you to detect, analyze and index up to 100 different faces in a single photograph (recall that the previous cap was 15). This means customers can now feed Amazon Rekognition a shot of a crowd of people and get the information in return regarding the demographics and sentiments of all the faces detected.

Yes. You take a group photo or an image at crowded public locations such as airports and department stores, and Amazon Rekognition will tell you what emotions the detected faces are displaying. Too good to be true!

On the large picture, Image Rekognition gives AWS a new shot in their repertoire. As more and more image content move into the internet, systems like Rekognition can help keep the customers glued to the cloud platforms, engaging businesses for longer periods of time. This is why Rekognition can further boost Amazon’s cloud business.

To get started with Text in Image, Face Search and Face Detection, you can download the latest SDK or simply log in to the Amazon Rekognition Console. For any further information, refer to the Amazon Rekognition documentation.


Subscribe to the weekly Packt Hub newsletter. We'll send you the results of our AI Now Survey, featuring data and insights from across the tech landscape.

LEAVE A REPLY

Please enter your comment!
Please enter your name here