Locating people in real-time video is now a reality, thanks to the launch of Amazon Rekognition Video, a new analysis service that allows the detection, tracking, recognition, extraction, and moderation of thousands of objects, faces, and content from a video.
The service builds on the Amazon Rekognition Image service launched last year and created by using deep learning neural network models based on the same technology that enables Amazon Prime Photos to analyse billions of images each day.
Launching the service at AWS re:Invent in Las Vegas on Wednesday, AWS CEO Andy Jassy said Amazon Rekognition Video also boasts the ability to detect inappropriate content on a website.
Rekognition Video can already isolate celebrity images, but is capable of being continually trained.
The new feature provides motion context, using deep learning to derive more complete insights about what activities are being performed in the video. For example, it can identify that there is a man, a car, and a tree in the video, as well as determine that the man in the video was running to the car.
Getting the video into the cloud is enabled by another feature launched on Wednesday, Amazon Kinesis Video Streams, which ingests and stores video, audio, and other time-encoded data from millions of camera devices without having to set up or run infrastructure.
"Kinesis Video Streams accepts your incoming streams, stores them durably and in encrypted form, creates time-based indexes, and enables the creation of vision-enabled applications," AWS explains in a blog post.
With reports that AWS was planning on launching a translation service for developers back in June, Jassy lifted the lid off the company's new AWS Transcribe feature, boasting long-form automatic speech recognition.
According to Jassy, the new service provides the automatic conversion of speech from video into accurate grammatically correct text, starting with English and Spanish.
He said unlike other transcription services in the market, AWS Transcribe will not show up as one long stream of text; rather through the use of machine learning, the feature will add in grammatical modifications. It can also be trained, with the user able to add their own custom libraries and vocabularies.
Amazon Translate, also unveiled at AWS re:Invent, automatically translates text from audio between languages in real-time.
Jassy on Wednesday also introduced Comprehend, a fully-managed natural language processing (NLP) service. It can analyse documents, social networking posts, and other content and give customers highly-accurate information.
Comprehend offers analysis in four categories for now: Language detection; entity categorisation, that is whether content includes people, places, brands, etc; sentiment analysis; and key phrase extraction.
The service "completely changes the meaning you can get from your content," Jassy said.