Local governments maintain thousands of infrastructure assets across the right of way, but the data behind those assets is often incomplete, inconsistent, or years out of date. Updating it typically means manual field surveys, disconnected spreadsheets, and time that GIS teams do not have. This session looks at how computer vision and machine learning models trained on street-level imagery and LiDAR point clouds can automate the production of structured, geospatially referenced infrastructure data: pavement condition scores and distress classifications, sign inventories coded to MUTCD standards, pavement markings, and right of way assets like hydrants, poles, and curbing. The session walks through the full data pipeline from sensor to GIS-ready outputs, including how imagery is collected, how models detect and classify assets with spatial precision, and how the resulting datasets integrate with existing tools through standard export formats. For GIS professionals supporting public works teams, the practical question is whether this approach fits into current workflows and what it changes about the quality, coverage, and frequency of the data they manage. This session addresses that directly.
