Official PyTorch models of "Raster2Seq: Polygon Sequence Generation for Floorplan Reconstruction" (SIGGRAPH'26)

Hao Phung     Hadar Averbuch-Elor

Cornell University  

[Page]    [Paper]    [Code]   

This repository hosts Raster2Seq's PyTorch checkpoints for floorplan reconstruction. Each checkpoint is stored in its own subfolder so users can download all checkpoints or only the subfolder they need.

License

Checkpoint releases distributed with this notice are offered under Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0). See LICENSE for scope, attribution, and commercial permission inquiries.

The Raster2Seq source code is separately licensed under MIT.

Available Checkpoints

Checkpoint key	Dataset	RoomF1	Subfolder
s3d-bw	Structured3D-B	99.6	s3d-bw/
cubicasa5k	CubiCasa5K	88.7	cubicasa5k/
raster2graph	Raster2Graph	97.0	raster2graph/
raster2graph-512	Raster2Graph-512	98.1	Raster2Graph-512/
s3d-density	Structured3D-DensityMap	99.1	s3d-density/

Download All Checkpoints

from huggingface_hub import snapshot_download

local_repo = snapshot_download(repo_id="haopt/Raster2Seq")

Download One Checkpoint Subfolder

from huggingface_hub import snapshot_download

local_repo = snapshot_download(
    repo_id="haopt/Raster2Seq",
    allow_patterns=["config.json", "Raster2Graph-512/*"],
)

Raster2Seq Helper

With the Raster2Seq codebase, users can load by alias:

python eval.py --checkpoint hf:cubicasa5k ...

or download directly:

from raster2seq_hub import download_checkpoint

ckpt_path = download_checkpoint("cubicasa5k")

Please CITE our paper and give us a :star: whenever this repository is used to help produce published results or incorporated into other software.

@inproceedings{phung2026raster2seq,
   title={Raster2Seq: Polygon Sequence Generation for Floorplan Reconstruction},
   author={Phung, Hao and Averbuch-Elor, Hadar},
   booktitle={Special Interest Group on Computer Graphics and Interactive Techniques Conference Conference Papers},
   year= {2026},
}