<i>DeepMapper:</i> A GeoAI Approach to Automate the VGI Mapping Workflow

AI-powered mapping technologies are uniquely positioned to address the challenges of keeping online maps up-to-date in dynamic urban environments. Modern Geospatial Artificial Intelligence (GeoAI) approaches, particularly those utilizing advanced Deep Learning (DL) techniques, have shown significant promise in automating aspects of the Volunteered Geographic Information (VGI) map updating process. This paper introduces DeepMapper, an open-source, GeoAI-based mapping platform designed to automate the VGI mapping workflow with minimal human intervention. By applying Generative Adversarial Networks (GANs) to freely available spatial datasets like OpenStreetMap, DeepMapper enhances the accuracy and completeness of VGI maps. As a web-based tool, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DeepMapper</i> is a web-based mapping tool that streamlines/automates much of the traditional VGI map updating workflow. Experiments demonstrate that DeepMapper achieves an accuracy of 92.8%, precision of 90.3%, and recall of 92.0%. The platform also shows potential for testing new GeoAI models, facilitating continuous improvements in how contemporary VGI maps are created and maintained. Despite its successes, challenges remain, particularly in adapting DeepMapper to real-world settings and different architectural styles. Further work will explore the integration of diffusion models for building detection and address non-technical challenges related to geometric accuracy and the adoption of AI-based mapping by communities like OpenStreetMap.