Awesome

House-GAN++

Code and instructions for our paper: House-GAN++: Generative Adversarial Layout Refinement Network towards Intelligent Computational Agent for Professional Architects, CVPR 2021. Project website.

Data

alt text We have used the RPLAN dataset, which offers 60k vector-graphics floorplans designed by professional architects. Qualitative and quantitative evaluations based on the three standard metrics (i.e., realism, diversity, and compatibility) in the literature demonstrate that the proposed system outperforms the current-state-of-the-art by a large margin.<br/> <br/>

Demo

Please check out our live demo.

Running pretrained models

See requirements.txt for checking the dependencies before running the code

For running a pretrained model check out the following steps:

Run python test.py.
Check out the results in output folder.

Training models

Download the raw RPLAN dataset.
Run this script for processing the dataset and extracting JSON files.
The extracted JSON files serves directly as input to our dataloader.

Citation

Please consider citing our work.

@inproceedings{nauata2021house,
  title={House-GAN++: Generative Adversarial Layout Refinement Network towards Intelligent Computational Agent for Professional Architects},
  author={Nauata, Nelson and Hosseini, Sepidehsadat and Chang, Kai-Hung and Chu, Hang and Cheng, Chin-Yi and Furukawa, Yasutaka},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={13632--13641},
  year={2021}
}

Contact

If you have any question, feel free to contact me at nnauata@sfu.ca.

Acknowledgement

This research is partially supported by NSERC Discovery Grants, NSERC Discovery Grants Accelerator Supplements, DND/NSERC Discovery Grant Supplement, and Autodesk. We would like to thank architects and students for participating in our user study.