[Discussion] Street View Insights API & Edge AI Pipelines

[grave mason]

Really interested in the Street View Insights mentioned earlier. For those of us building Edge AI agents, the "Grounding with Google Maps" feature in Gemini is great for high-level queries, but we often need raw, lower-level access for custom perception pipelines.

Does anyone know if the new Insights API exposes any structured feature embeddings (like road signs or asset vectors) directly? Or is it strictly a "black box" query service?

Trying to figure out if I can plug this into a local MediaPipe / TFLite workflow, or if I'm stuck with cloud latency.

[heavy rune]

The Street View Insights service is currently a Restricted GA analytics offering, not a real-time request-response API. The service pushes data into your Google Cloud Project for offline analysis.

If your Edge AI agents require a real-time local perception of their immediate surroundings, Street View Insights won't act as a live sensor fed. However, if your agents need to "know the world" beforehand (e.g., an inventory of all road signs in a city), this service provides the raw image snippets and structured data to build that foundational knowledge in the cloud.