LAS VEGAS — Amazon Web Services added several new items to its suite of machine-learning driven services Wednesday at AWS re:Invent 2017, including the extension of its Amazon Rekognition image-detection technology to video.
AWS CEO Andy Jassy walked attendees through the new services, which AWS customers can run on data stored in their S3 storage buckets to unlock new insights. Perhaps the most interesting was Rekognition video, which uses machine learning to detect inappropriate content or track objects or people across frames, he said.
Jassy highlighted two examples of how this service could be used in smart home camera technology. Video Rekognition could detect and scan packages as they are delivered to your doorstep, which could help detect theft or fraud, or it could open your garage door when it scans and correctly identifies your license plate.
Video publishers can also use it to block certain videos from their websites or identify celebrities in videos, Jassy said. And if you’re worried about how you’re going to get that much video into AWS, the company announced a version of its Kinesis data-streaming service just for video.
“The service will continue to get better every month just with the sheer volume of video we have internally, as well as public datasets,” he said.
Two other new services — Amazon Transcribe and Amazon Translate — will do pretty much what you might expect of services bearing those names. And another new service called Amazon Comprehend will be able to analyze all of your text files and pick out interesting nuggets or analyze social-media posts for mentions of your product or brand, a service Salesforce also provides to its customers.