AWS: New algorithm knows difference between a dog, a human and a package on live video streams

Amazon Rekognition Streaming Video Events uses machine learning to send custom alerts from home security cameras.

AWS announced an update to its video monitoring service that can tell the difference between a pet, a person and a package on home security camera devices. Amazon Rekognition Streaming Video Events uses machine learning to detect specific objects in video streams from home security devices.

The new service can send customized alerts such as “pet detected in the backyard” and is now in general availability.

Amazon Rekognition Streaming Video Events works with Amazon Kinesis Video Streams to detect objects from live video streams. Camera manufacturers and service providers can minimize false alerts from camera motion events by sending real-time notifications only when a desired object is detected in the video frame. Service providers also can use Amazon Rekognition streaming video APIs to send an alert for objects that are relevant for their customer, successfully adjust the duration of the video to process per motion event and define specific areas within the frame for analysis.

When Amazon Rekognition starts processing the video stream, it sends a notification as soon as a desired object is detected. This notification includes the object detected, the bounding box, the time stamp and a link to the specified Amazon Simple Storage Service (Amazon S3) bucket with the zoomed-in image of the object detected.

Amazon Rekognition automatically encrypts data at rest using AWS Key Management Service and in transit using the industry-standard Transport Layer Security protocol.

SEE: AI investments soared in 2021, but big problems remain

According to AWS, this service lowers the costs of value-added video analytics for service providers by providing a low-cost, low-latency and fully managed ML service. Camera manufacturers and service providers can add new features and enhance existing capabilities with video analysis on live video streams, including:

1. Integrate Kinesis video streams: Stream live video from video doorbells and indoor and outdoor cameras to AWS and integrate with new or existing Kinesis streams with Amazon Rekognition for live analysis.
2. Specify video duration: Control how much video needed to process per motion events, from 1–120 seconds (with a default of 10 seconds). When motion is detected, Amazon Rekognition starts analyzing video from the relevant Kinesis video stream for the specific duration to provide the flexibility to manage ML inference costs.
3. Choose relevant objects: Choose one or more objects for detection in live video streams to minimize false alerts and send notifications only when desired objects are detected in the video frame.
4. Customize event notifications: Specify when and how often Amazon’s Simple Notification Service sends event notifications to send smart alerts to users.
5. Send motion detection trigger notifications: Amazon Rekognition sends a notification to their private SNS topic when the desired object is detected.
6. Integrate with Alexa or other voice assistants: Service providers can connect these notifications with Alexa Smart Home skills to enable Alexa announcements or audio announcements from Echo devices, such as “Package detected at the front door.”

Amazon Rekognition Video Streaming Events is generally available to AWS customers in U.S. East (N. Virginia), U.S. East (Ohio), U.S. West (Oregon), Europe (Ireland) and Asia Pacific (Mumbai) Regions, with availability in additional regions in the coming months.