Scaling Open 3D City Models: Implementation and Validation of an AI-driven Automated Generation Tool “AI City Model Maker beta version”

Abstract. This study presents an AI-based tool for the automatic generation and updating of 3D city models to support the expansion of Project PLATEAU in Japan. High-quality models at Level of Detail (LOD) 2 or higher are essential for applications such as urban planning and disaster management, but their production currently relies on labor-intensive manual processes. To address this issue, the proposed system, AI City Model Maker (Beta Version), integrates multiple AI techniques to generate buildings, roads, city furniture, and vegetation models from heterogeneous geospatial data, including imagery, DEM, and point clouds.
The tool is designed to comply with stringent quality requirements, such as positional accuracy and logical consistency. However, fully automated processing alone remains insufficient to meet these standards, particularly when using commonly available datasets. Therefore, a hybrid workflow combining automated processing and human verification is adopted, with a target cost reduction of 30-50%. The beta version has been tested in multiple real-world environments, demonstrating promising performance, especially for road and urban object modeling. The results indicate the potential to significantly improve efficiency while maintaining practical usability for municipal applications.