What semantics in VGI and theSocial Web can better support generating richer building models for sustainable cities?

Abstract

For urban GeoAI deep-learning processing pipelines, the integration of user-generated content, Volunteered Geographic Information (VGI), and crowdsourced OpenStreetMap (OSM) data becomes more and more relevant to reflect better on the ground and in-situ understanding of citizens versus geometry-focussed and often quite mechanistic engineering approaches. This inclusion catalyzes enriched semantics of digital building models, particularly enhancing attributes such as functions and types. By harnessing the collective intelligence of users, these models become more comprehensive and reflective of dynamic urban landscapes. This synergy is vital for sustainable smart cities and urban planning applications. But is this enough? Based on examples, the talk will review and discuss research questions and recent approaches that go beyond classical approaches for crowdsourced labelling for deep-learning training data. While we will discuss some recent crowdsourcing infrastructures and tools for deriving such information (e.g. ohsome2label, MapSwipe, mapSwipe Web, SketchMapTool, to name a few), the focus will be on innovative aspects such as deriving information about cultural and social significance, accessibility and sentiments, the local businesses ecosystem potential sustainability practices, public art or aesthetic values or related to the locations, and especially including temporal dynamics, adaptive reuse and evolution. This would allow for a more holistic understanding, support informed decision-making in urban planning, and enable new smart sustainable city applications.