Make API call to Machine Learning module on IoT Edge

Question

I've been able to push down a machine learning module to an IoT edge device as per the below Microsoft tutorial. I am also able to get ML predictions from routing data from the tempSensor simulated data module to the ML module which is great.

https://docs.microsoft.com/en-gb/azure/iot-edge/tutorial-deploy-machine-learning

What i would like to do is use the Machine Learning module as a web service on the IoT Edge device - is there a way to target this module using REST etc? Something similar to the below, although i'm pushing down the ML model as a module to the IoT edge device not as a web service.

https://docs.microsoft.com/en-gb/azure/machine-learning/preview/model-management-service-deploy

EDIT: I'm not looking to deploy a ML image as a web service like the following link using "az ml service create" etc. I would like to deploy the ML image as a module via IoT Edge management and still access it via REST API. https://docs.microsoft.com/en-gb/azure/machine-learning/preview/model-management-service-deploy

Cheers, Com

All Azure Machine Learning models containerized as Docker-based web services can also run on Azure IoT Edge devices.Have you can refered to Deploying a Machine Learning Model as a web service? — Michael Xu - MSFT
Thanks for the reply. Yes i have. The workflow here uses this instead: az ml service run realtime -i <service id> -d "etc". I can use this on the Edge device (I assume?) after i set it up as a local environment. This would pull the docker image from the repository directly. The downside of this i believe is you cannot use it in the routing between modules which is in the IoT Edge architecture. If i use the workflow in pushing a module down via IoT Edge management (docs.microsoft.com/en-gb/azure/iot-edge/…), can i still access the REST service? — Commio

Ted W. Way Ted W. Way · Accepted Answer · 2018-02-06T06:50:26

Yes, the REST API for scoring your data in the Docker container on the Azure IOT Edge device can be called directly. Here is more information on consuming a web service: https://docs.microsoft.com/en-gb/azure/machine-learning/preview/model-management-consumption

If you are interested in creating your own custom AI/ML models, I would recommend going through the Azure ML Iris tutorial: https://docs.microsoft.com/en-gb/azure/machine-learning/preview/tutorial-classifying-iris-part-1

This will walk you through training a model and operationalizing it to a Docker container. You can inference by calling the REST API with your data whether the container is deployed to the cloud or on the edge.

Make API call to Machine Learning module on IoT Edge

1 Answers