Using deep learning to identify (urban) form and function in satellite imagery - the case of Great Britain

In this paper, we introduce a framework to leverage satellite data through AI to build rich representations of urban form and function. We use a concept of Spatial signatures to characterise predominantly urban environment into data-driven classes based on both form and function composed of a large number of input data sources. Consequently, we explore the ability of Sentinel 2 satellite imagery and state-of-the-art AI models to capture the same classification of space using a single, regularly updated data source, including the conceptual questions of relationship between granular signature geometry and rigid raster grid of satellite data.