Image Machine Learning has been recently applied to self-driving cars, automated airport border security gates, and road surveying. Over the past 5-10 years, deep neural network machine learning has been successfully applied to many image recognition problems, and in some areas surpassed human-level accuracy. This talk will present an overview of the research and development work that I led as part of a Spatial and Image Learning team in machine learning application for surveying road signage and road surfaces (in collaboration with NZTA). The team has overcome some of the shortcomings in applying the state-of-the-art machine learning algorithm, including successfully outlining thin objects (‘needles’) in 3D Lidar data for NZ roads.