Lopata Hall, Room 101
"Raster-to-Vector: Revisiting Floorplan Transformation"
Adviser: Yasutaka Furukawa
We address the problem of converting a rasterized floor plan image into a vector-graphics representation using a learning-based approach. A neural architecture first transforms a rasterized image to a set of junctions that represent low-level geometric and semantic information. Integer programming is then formulated to aggregate junctions into a set of simple primitives to produce a vectorized floorpan, while ensuring a topologically and geometrically consistent result. Our algorithm significantly outperforms existing methods and achieves around 90% precision and recall, getting to the range of production-ready performance. The vector representation allows 3D model popup for better indoor scene visualization, direct model manipulation, and floorplan document analysis.