A multibranch and multiattention framework for floor plan segmentation

Abstract

Automated floor plan analysis is crucial in architecture, urban planning, and interior design. Floor plan segmentation is a foundational step for tasks such as surface area estimation and three‐dimensional building reconstruction. However, automatic semantic segmentation of floor plan images faces unique challenges, including high interclass similarity, ambiguous room boundaries, and varying floor plan styles. We introduce a novel multibranch and multiattention framework for deep floor plan segmentation, explicitly designed to handle the challenges of interclass similarity, ambiguous room boundaries, and diverse architectural styles. Our method leverages intrabranch channel attention and cross‐branch positional attention to refine both boundary recognition and room‐type segmentation, significantly enhancing robustness and accuracy across multiple datasets. Through extensive experiments on the raster‐to‐vector (R2V) and R3D datasets, we demonstrate how our approach sets a new state‐of‐the‐art for floor plan segmentation, outperforming general‐purpose and specialized models alike.