VectorFloorSeg: Two-Stream Graph Attention Network for Vectorized Roughcast Floorplan Segmentation

Bingchen Yang

Abstract:

Vector graphics (VG) are ubiquitous in industrial de
signs. In this paper, we address semantic segmentation
of a typical VG, i.e., roughcast floorplans with bare wall
structures, whose output can be directly used for further
applications like interior furnishing and room space mod
eling. Previous semantic segmentation works mostly pro
cess well-decorated floorplans in raster images and usu
ally yield aliased boundaries and outlier fragments in seg
mented rooms, due to pixel-level segmentation that ignores
the regular elements (e.g. line segments) in vector floor
plans. To overcome these issues, we propose to fully uti
lize the regular elements in vector floorplans for more in
tegral segmentation. Our pipeline predicts room segmen
tation from vector floorplans by dually classifying line seg
ments as room boundaries, and regions partitioned by line
segments as room segments. To fully exploit the structural
relationships between lines and regions, we use two-stream
graph neural networks to process the line segments and par
titioned regions respectively, and devise a novel modulated
graph attention layer to fuse the heterogeneous information
from one stream to the other. Extensive experiments show
that by directly operating on vector floorplans, we outper
form image-based methods in both mIoU and mAcc. In ad
dition, we propose a new metric that captures room integrity
and boundary regularity, which confirms that our method
produces much more regular segmentations.