By Manush Bhatt, Rajesh Kalyanam (contributor), Gen Nishida (contributor), Liu He (contributor), Christopher K May (contributor), Dev Niyogi (contributor), Daniel Gerardo Aliaga (contributor)

Construct a 3-D building model (reconstruction) given a picture of a building

Launch Tool

This tool version is unpublished and cannot be run. If you would like to have this version staged, you can put a request through HUB Support.

Archive Version 1.04
Published on 15 Jan 2020, unpublished on 27 Mar 2023 All versions

This tool is closed source.



Published on


Procedural modeling is a popular approach to synthesize urban environments but requires writing suitably parameterized grammars. In this tool, we automate the generation of procedural buildings by taking a photograph as example input. Our system does not aim at an exact reproduction of a building, but rather at capturing its overall shape, the layout of its facade, and the style of its windows. To do so, we decompose the problem into logical stages (mass, façade, windows) and treat each stage with a common methodology that consists of simplifying the input to make it amenable to analysis by deep networks trained with synthetic data, and refining the output with custom optimizations. The resulting pipeline can generate a diversity of procedural buildings with no user intervention.


NSF Award 1904444 (also supported by Internet2 and Google Cloud Platform), NSF CSSI 1835739, NSF CBET 1250232, and NSF IIS 1816514

Cite this work

Researchers should cite this work as follows:

  • Manush Bhatt; Rajesh Kalyanam; Gen Nishida; Liu He; Christopher K May; Dev Niyogi; Daniel Gerardo Aliaga (2019), "Photo2Building,"

    BibTex | EndNote