Visualisation of the randomly generated buildings with my procedural modelling engine Random3Dcity. The synthetic data is generated in a procedure that regards a multitude of rules and constraints to derive a relatively realistic set of buildings. This engine creates the models in multiple LODs and multiple geometric references, and exports them in CityGML.
The methodology is described in the following paper: doi.org/10.5194/isprs-annals-III-4-W1-51-2016
The data is used for various analyses in my PhD, and it is publicly available on my website:
Workflow: Python Random3DCity (generate buildings) -> CityGML data realisation -> CityGML2OBJs tool -> OBJ -> Blender.
The video shows a version with 10000 buildings.
CityGML to OBJ converter: