Publishing an ML model using MMS
Now that you have seen ML model-based object detection happening at the Edge device node, we will use MMS to publish an updated ML model. In this example, to illustrate the deployment of the new model, we have created another version of the same model where some of the objects are wrongly labeled – for example, TV with radio, and so on. We will use two different models with two different metadata model publish files.
Important note
The focus of this example application is not the ML model itself but the delivery of the ML model to the edge nodes asynchronously using MMS.
To receive the MMS-published ML model and to make the model available to the application inference pipeline, this application deploys a standalone mms service. In this application design, the mms and infer services share a common directory structure where the newly received ML model is stored and then later retrieved by the infer service to reinitialize the inference...
