This demo walks through the professional services ML Journey to create, train and deploy a GCP AI Platform Unified Auto ML Image Segmentation model. The demo that shows how, one could leverage this model to do batch predictions on over 2 million STREAM:RASTER tiles, store the results in BigQuery and render the results as a layer in a GIS application using Cloud Run.
- Proves out doing ML at scale, in the cloud, for Woolpert
- Demonstrates an ML data pipeline
- Demonstrates an ML client application (opportunity for productizing)
- Demonstrates leveraging a data pipeline from BigQuery to Cloud Run
- Possible bonus - leverage existing model and train with additional labels
Where are the trampolines? The demo is still available, though only use it if you know how to explain/reconcile it’s poor results and lack of good training data. It’s over here.