-
Interested in building your first machine learning (ML) inference application at the edge and not sure where to start? With Verizon 5G Edge, connect your mobile 5G devices to compute infrastructure at the network edge to process images in real time. By storing your model artifacts in the parent region, leverage Wavelength Zones to decouple your model training from inference engine for immediate application response times. In this example, sending an image from a 5G device to 5G Edge boosts the delivery speeds by 25x with 5G UWB connectivity. To learn more about image classification and distributed ML at the network edge, contact our professional services team here and choose “5G Edge Computing – MEC” in the Area of Interest drop-down box.
Architecture overview
Client node connects to Amazon Cognito℠ to authenticate and generate tokens to access APIs
Client node requests session setup; edge worker node is selected; session and inference servers for the session are orchestrated; session setup completed
Feed node acquires and sends high-resolution video to inference server
Client node receives video stream with ML inference metadata and alerts
Session server uploads the inference metadata and images to Amazon® Simple Storage Service (S3) for review, reports and re-training
Client node initiates session teardown; session metadata, statistics uploaded to database
