The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Download
Share
Publications Copernicus
Download
Citation
Share
Articles | Volume L-4/W1-2026
https://doi.org/10.5194/isprs-archives-L-4-W1-2026-143-2026
https://doi.org/10.5194/isprs-archives-L-4-W1-2026-143-2026
29 Aug 2026
 | 29 Aug 2026

From CityGML to Game Engine: A Reproducible Open-Source GIS Pipeline for Interactive 3D Urban Environments Using PLATEAU, Blender and Godot

Léo Martial, Hiroyuki Kubono, and Shino Miura

Keywords: CityGML, PLATEAU, Godot Engine, Open-Source GIS, 3D City Models, Reproducible Pipeline

Abstract. Open 3D city-model datasets are increasingly available worldwide through the CityGML standard, and Japan’s Project PLATEAU is among the largest national open initiatives of its kind. Yet official software development kits exist only for the proprietary Unity and Unreal engines, and no documented route connects PLATEAU data to an open-source interactive environment. This paper presents and evaluates a fully open-source geospatial pipeline that transforms CityGML data into a lightweight, interactive, real-time 3D application suitable for the low-specification hardware typically found in schools and community centres. The pipeline relies exclusively on free and open-source software, using Blender for geospatial processing and 3D optimisation and the Godot Engine for real-time rendering and interaction. We document the pipeline at the level of detail required for reproduction, including the PLATEAU data model and tiling scheme, the reprojection from the native geographic frame (EPSG:6697) to a metric local frame, the geometry-repair and level-of-detail decisions, and the automatic generation of the collision geometry that a game engine requires but a GIS does not. The pipeline was validated as a deployment stress test in a participatory urban-design workshop with approximately fifty elementary-school students in Yokohama, during which 1,192 assets were placed without system failure on six standard classroom computers. We identify Blender as the indispensable bridge component and the collision layer as the technical linchpin, and we argue that the present manual workflow, while reproducible, is still too complex for non-experts; we therefore outline a realistic turnkey, automated successor.

Share