A sign inventory that keeps itself current
An end-to-end computer vision pipeline turning roadway imagery into a located, classified, condition-rated sign inventory on a map — kept current by new…
The decision
Which signs need replacing, and where are they?
The problem
The inventory was a survey crew that drove the network once, some years ago, and a spreadsheet that was accurate on the day it was finished. Everything since — a sign knocked down, a sign replaced, a tree grown across one — is missing from it, and the maintenance schedule and replacement budget are wrong by however far it has drifted.
Outcome
A production pipeline converting roadway imagery into a located, classified and condition-rated sign inventory, published as a map layer with per-asset evidence, and updated by the next collection pass instead of by the next survey.