Research map / Point clouds / Point cloud classification
PolyRoom: Room-aware Transformer for Floorplan Reconstruction
Yuzhou Liu, Lingjie Zhu, Xiaodong Ma, Hanqiao Ye, Xiang Gao, Xianwei Zheng, Shuhan Shen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add uniform sampling floorplan representation (ablation) | PolyRoom · Structured3D test | Corner precision: 78.2 -> 94.6 | Improved |
| add room-aware query initialization from instance segmentation (ablation) | PolyRoom · Structured3D test | Corner precision: 89.3 -> 94.6 | Improved |
| replace vanilla self-attention with room-aware self-attention (ablation) | PolyRoom · Structured3D test | Corner precision: 94.1 -> 94.6 | Improved |
| room-aware self-attention instead of vanilla (ablation) | PolyRoom · Structured3D test | GPU memory: 18707M -> 11957M | Improved |
| add vertex selection (ablation) | PolyRoom · Structured3D test | Corner precision: 81.4 -> 94.6 | Improved |
| add angle loss (ablation) | PolyRoom · Structured3D test | Angle precision: 85.0 -> 89.3 | Improved |
| add polygonization in vertex selection (ablation) | PolyRoom · Structured3D test | Corner precision: 92.6 -> 94.6 | Improved |
| combine intra-room and inter-room attention rather than either alone (ablation) | PolyRoom · Structured3D test | Corner precision: 93.3 -> 94.6 | Improved |
| increase vertex number from 40 to 50 (ablation) | PolyRoom · Structured3D test | Corner precision: 94.6 -> 94.4 | Hurt |
| increase decoder layers from 6 to 8 (ablation) | PolyRoom · Structured3D test | Room precision: 98.9 -> 98.7 | No clear effect |
| use stronger instance segmentation network for initialization (ablation) | PolyRoom · Structured3D test | Corner precision: 92.6 -> 94.6 | Improved |
| use PolyRoom instead of RoomFormer in cross-dataset transfer | PolyRoom · Structured3D to SceneCAD | Corner F1: 60.3 -> 78.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SLIBO-Net — SLIBO-Net: Floorplan Reconstruction via Slicing Box Representation with Local Geometry Regularization (NeurIPS 2023) · better on 2 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Connecting the Dots: Floorplan Reconstruction Using Two-Level Queries (CVPR 2023 · similarity 0.93)
- CAGE: Continuity-Aware edGE Network Unlocks Robust Floorplan Reconstruction (NeurIPS 2025 · similarity 0.90)
- SLIBO-Net: Floorplan Reconstruction via Slicing Box Representation with Local Geometry Regularization (NeurIPS 2023 · similarity 0.90)
- FRI-Net: Floorplan Reconstruction via Room-wise Implicit Representation (ECCV 2024 · similarity 0.90)
- Floor-SP: Inverse CAD for Floorplans by Sequential Room-Wise Shortest Path (ICCV 2019 · similarity 0.89)
- HouseLayout3D: A Benchmark and Training-free Baseline for 3D Layout Estimation in the Wild (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
