This repository contains the code for Phone2Proc. It performs conditional procedural generation to build a dataset of houses based on a specified RoomPlan USDZ file. The backend uses AI2-THOR and ProcTHOR. The generated houses are in the ProcTHOR house JSON format (see this tutorial on loading them into ProcTHOR).

To obtain RoomPlan USDZ files, you'll need to load a custom app onto your phone, scan your scene, and then export it to a USDZ.

See the this tutorial and the RoomPlan tutorial for more help.

To build the houses, install the requirements:

pip install -r requirements.txt
pip install --extra-index-url https://ai2thor-pypi.allenai.org ai2thor==0+391b3fae4d4cc026f1522e5acf60953560235971

Then run:

python3 dmain_wrapper.py

It'll generate houses based on the specified RoomPlan layout. You may need to go into main.py and change some constants, such as which USDZ file is being used.

Note that the RoomPlan scene must have all walls form an enclosed loop. Otherwise the generation will not work as it'll be underspecified how to generate the floorplan.

To cite Phone2Proc, please use the following entry:

@inproceedings{phone2proc,
  title={Phone2proc: Bringing robust robots into our chaotic world},
  author={Deitke, Matt and Hendrix, Rose and Farhadi, Ali and Ehsani, Kiana and Kembhavi, Aniruddha},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={9665--9675},
  year={2023}
}

AI2-THOR is an open-source project built by the PRIOR team at the Allen Institute for AI (AI2). AI2 is a non-profit institute with the mission to contribute to humanity through high-impact AI research and engineering.