The open source libraries for machine learning and feature detection have dramatically improved over the past two years. These tools have tremendous power to aid in OpenStreetMap’s goals, not by replacing mappers but by supercharging their efforts–helping them to work faster and produce more accurate results.

Development Seed has two years of experience developing and applying machine learning tools to imagery classification and feature identification. We’ll talk through the current state of machine learning and present a vision for building these tools into OpenStreetMap workflows. Examples include prioritization based on predicted features not already in OpenStreetMap, correction of errant traces, and fraud/vandalism detection.