M+E Daily

Amazon Web Services (AWS) provides a wide selection of artificial intelligence (AI) and machine learning (ML) services that are designed to fit all of a customer’s needs. During the online AWS Immersion Workshop on AI/ML Basics online event Sept. 14, the company presented a deep dive into the capabilities and use cases of one of those services: Amazon Rekognition.

Launched by the company in late 2020, Amazon Rekognition is a deep learning image and video classification service that makes it easy to detect, analyze and compare faces or objects for a wide variety of application needs.

“Our mission at AWS is to put machine learning in the hands of every developer,” according to Rhea Lingaiah solutions architect at AWS. “We want to be able to provide our customers access to machine learning and AI tools without needing prior knowledge or being a machine learning expert,” she pointed out.

And Amazon Rekognition “does just that,” she said, noting it’s a service that “requires no machine learning expertise to use.”

“With Amazon Rekognition, you can identify objects, people, text and scenes and images and video,” she said, noting it also “provides highly accurate facial analysis and facial search capabilities that you can use to detect, analyze and compare faces for wide use cases.”

Amazon Rekognition provides multiple application programming interfaces (APIs) for image and video analysis, she pointed out. “This returns various different types of visual metadata, depending on the API call that was made. So our customers can use Rekognition to identify objects, scenes and activities, perform face analysis, face search and face comparison and identify celebrities, detect unsafe content, detect text,” monitor live streams and more, she told viewers.

“So why should customers use Amazon Rekognition?” she asked rhetorically. The answer she gave: “It’s an out of the box solution that is ready to go for customers who don’t have ML expertise but need that image and video analysis for their business. Customers can easily integrate their applications with the APIs and utilize the custom labels feature if they want to be able to find objects and scenes and images specific to their business needs.”

There have been some “big use cases within the media industry” for Amazon Rekognition, she said, noting it has been used for live streams. For example, The New York Times used it during its royal wedding coverage to help identify guests, she said.

“Amazon Rekognition automatically extracts metadata, such as names of celebrities, scenes [and] activities,” she said, adding: “All of this enables our customers to create a searchable image and media library. You can also use Rekognition for social media. So when managing user-generated content, you can detect explicit content and create rules around what is appropriate for the culture and the demographic of your users.”

“Influencer marketing platforms have also used the image recognition to label searchable content and enable users to source influencers within seconds,” she told viewers.

There is also an object and scene detection feature, she said. “When you specify an image as an input, the service detects the objects and scenes in the image and returns them along with the percent confidence score for each object and scene,” she noted. The “confidence score is a value that quantifies the certainty of the model’s prediction,” she explained.

“Rekognition image operations can also return bounding box coordinates for the items that are detected in images,” she said, referring to the rectangular frame that you see on an image that “fully encompasses the face or object to locate the position in the image.”

Users can also extract text by using Rekognition, she said. “The text detection feature now supports seven new languages with Arabic, Russian, German, French, Italian, Portuguese and Spanish,” she pointed out.

Another use case for text detection is with automotive applications to read text on road and street signs and in public and safety transportation use cases, she added.