CRAB: Cuboid-based Representation for As-built BIM from Point Clouds

Haoran Xiong; Jing Wang; Shanshan Pan; Zihao Yan; Minglun Gong; Hui Huang

cs.CVOct 21, 2026

CRAB: Cuboid-based Representation for As-built BIM from Point Clouds

Authors: Haoran Xiong, Jing Wang, Shanshan Pan, Zihao Yan, Minglun Gong, Hui Huang

Organizations: Guangdong Provincial Key Laboratory of Visual Media and Multidimensional Intelligence, College of Computer Science and Software Engineering, Shenzhen University, China, Shenzhen · University of Guelph, Canada, Guelph

Abstract

As-built BIM reconstruction from point clouds demands semantic, volumetric, parametric, and editable building components, beyond 2D layouts or planar abstractions. Yet existing scan-to-BIM pipelines often rely on handcrafted rules, heuristic optimization, or manual correction, while learning-based methods fall short of recovering wall thickness, opening geometry, and component-level BIM structure. We introduce CRAB, a BIM-ready cuboid reconstruction framework that encodes walls, doors, windows, beams, and columns as compact architectural cuboids with IFC-compatible semantics and parameters. CRAB bridges point clouds and editable BIM by combining ProposalNet, which predicts coarse component cuboids, with AssemblyNet, a relation-aware Flow Matching model that refines geometry, predicts structural relationships, and enables automatic structural correction and human-guided editing. To address limited cuboid-level supervision, we extend Structured3D with component-wise cuboid annotations and construct cuboid-based procedural scenes. Experiments show CRAB surpasses floor plan and plane-based alternatives in component completeness, geometric accuracy, and generalization. For interior wall thickness estimation, CRAB achieves a mean error of 2.80 cm, with 79.50% of matched walls within 3 cm. The project page is available at https://vcc.tech/research/2026/CRAB.

Figures & tables

Figure 1. CRAB converts cluttered and incomplete indoor scans into compact, BIM-ready cuboid assemblies. By representing walls, doors, windows, beams, and columns as semantic cuboids with explicit geometric parameters and structural relations, CRAB supports accurate reconstruction, local editability, and downstream BIM conversion. The zoomed regions highlight challenging structures, including dense room partitions, non-uniform wall thicknesses, corner openings, and complex interior cores, while demonstrating precise wall-thickness recovery, consistent door-wall junctions, and robust reconstruction.

Dataset	Source	Scale	2D Floor Plan	3D Plane	Volumetric Comp.
ISPRS Benchmark ( 2017 )	Real-world	6	$\times$	$\times$	$\checkmark$
Structured3D ( 2020 )	Synthetic	3,500	$\checkmark$	$\checkmark$	$\times$
SceneCAD ( 2020 )	Real-world	1,151	$\checkmark$	$\times$	$\times$
ProcTHOR ( 2022 )	Synthetic	12,000	$\checkmark$	$\times$	$\times$
HSSD ( 2024 )	Synthetic	211	$\checkmark$	$\times$	$\times$
ASE ( 2024 )	Synthetic	100,000	$\checkmark$	$\checkmark$	$\times$

Table 1. Structural supervision in indoor building datasets. We compare whether each dataset provides 2D floor plan, structured 3D plane, and volumetric component-level annotations for BIM-oriented reconstruction.

Figure 2. Representative samples from CrabSet , combining S3D-derived residential and procedurally synthesized scenes for cuboid-level supervision. The synthesized subsets vary boundaries, scales, and corridor-centered layouts, covering non-uniform wall thicknesses, non-Manhattan contours, service shafts, and diverse circulation patterns.

Figure 3. Overview of CRAB . Given a raw indoor point cloud, ProposalNet encodes the scan and autoregressively predicts an initial set of semantic cuboids for walls, doors, and windows. AssemblyNet represents each proposal with centerline-based cuboid tokens augmented by semantic, relation, and edit-control flags, and refines them through intra-cuboid, inter-cuboid, and cuboid–point attention. A velocity head performs relation-aware flow-matching refinement of cuboid geometry, while a relation head infers structural dependencies such as wall adjacency and door/window affiliation. The refined cuboids and relations are finally instantiated as editable BIM-ready components, converting unstructured scans into component-level architectural models.

Method	Rep.	Exterior Wall	Interior Wall					Door		Window
Method	Rep.	Match-F1 ( $\uparrow$ )	Match-F1 ( $\uparrow$ )	CL-Offset ( $\downarrow$ )	T-MAE ( $\downarrow$ )	T-Acc@3 ( $\uparrow$ )	T-Acc@5 ( $\uparrow$ )	Match-F1 ( $\uparrow$ )	H-MAE ( $\downarrow$ )	Match-F1 ( $\uparrow$ )	H-MAE ( $\downarrow$ )
Cloud2BIM ( 2025 )	Volumetric	4.41%	62.76%	–	–	–	–	15.12%	–	4.52%	–
RoomFormer ( 2023 )	Floor Plan	84.48%	85.78%	5.34	6.19	27.79%	48.54%	77.05%	74.94	64.82%	120.84
CAGE ( 2025a )	Floor Plan	91.14%	89.74%	3.68	5.31	33.75%	58.11%	–	–	–	–
SceneScript ∗	Plane	44.40%	52.59%	6.02	9.24	10.73%	16.26%	47.83%	22.64	38.71%	22.01
SpatialLM ( 2025 )	Plane	92.40%	91.48%	3.10	3.89	24.98%	77.74%	89.03%	77.36	77.90%	122.86

Table 2. Quantitative comparison of representation choices on Structured3D. CRAB improves both component recovery and BIM-critical volumetric attributes, especially wall thickness and opening height, over floor plan and plane-based alternatives. Best and second-best results are highlighted in green and blue.

Figure 4. Qualitative comparison with a plane-based representation. Zoomed views compare the same local structures in GT, CRAB , and SpatialLM. CRAB ’s volumetric cuboids better preserve wall thickness, component extent, and opening-wall consistency than the plane-based representation.

Method	Rep.	2D Layout									3D Plane
		Room			Corner			Angle			IoU 2D @0.25				IoU 2D @0.5
		Prec.	Rec.	F1	Prec.	Rec.	F1	Prec.	Rec.	F1	Avg.	Wall	Door	Window	Avg.	Wall	Door	Window
RoomFormer ( 2023 )	Floor Plan	98.7	96.7	97.3	89.2	85.3	87.2	83.0	79.6	81.3	83.4	86.0	85.1	78.9	81.4	84.6	83.1	76.6
CAGE ( 2025a )	Floor Plan	99.6	98.7	99.1	95.0	88.6	91.7	92.5	86.4	89.3	–	89.3	–	–	–	86.4	–	–
SceneScript	Plane	81.3	45.3	58.2	71.1	29.8	42.0	68.4	28.6	40.4	90.4	92.9	92.8	85.5	89.2	91.9	92.5	83.3
SpatialLM ( 2025 )	Plane	99.3	97.2	98.2	94.5	89.5	92.0	93.7	88.7	91.1	92.2	93.5	94.1	88.9	91.1	92.4	93.4	87.5

Table 3. Quantitative comparison of 2D layout and 3D plane quality on Structured3D. The results show that CRAB’s gains in volumetric reconstruction do not come at the expense of layout or plane quality. Best and second-best results are highlighted in green and blue.

Figure 5. AssemblyNet refinement under simulated wall editing errors. The three rows show the ground truth, the perturbed input, and the refined result, respectively. Starting from the noisy wall highlighted in yellow, plane-based refinement can drift toward nearby structures, while cuboid-based refinement better preserves and recovers the target volumetric component.

Method	Exterior Wall	Interior Wall		Door	Window
Method	Match-F1 ( $\uparrow$ )	Match-F1 ( $\uparrow$ )	T-Acc@3 ( $\uparrow$ )	Match-F1 ( $\uparrow$ )	Match-F1 ( $\uparrow$ )
AssemblyNet (Plane)	87.94%	87.31%	20.45%	88.02%	76.44%
AssemblyNet (Cuboid)	94.42%	92.73%	81.27%	92.20%	89.76%

Table 4. Comparison of full-scene refinement with plane-based and cuboid-based AssemblyNet, using planes from SpatialLM and cuboids from ProposalNet, respectively. Cuboid-based AssemblyNet improves the initial reconstruction results, whereas plane-based AssemblyNet degrades them.

Figure 6. BIM conversion from cuboids and relations. The same cuboid-relation graph supports different junction assignments: Corner-Joint BIM uses explicit corner joints, Cross-Joint BIM uses explicit cross joints, and Wall-Integrated BIM merges junction regions into connected wall elements.

Method	Exterior Wall	Interior Wall		Door	Window
Method	Match-F1 ( $\uparrow$ )	Match-F1 ( $\uparrow$ )	T-Acc@3 ( $\uparrow$ )	Match-F1 ( $\uparrow$ )	Match-F1 ( $\uparrow$ )
Oriented Box	92.59%	92.06%	71.74%	92.20%	88.94%
Centerline Cuboid (Ours)	93.78%	92.57%	79.50%	91.14%	88.97%

Table 5. Ablation of cuboid parameterization. Oriented boxes perform well on compact components such as doors and windows, but are less suitable for long and thin walls, where longitudinal extent and thickness play strongly asymmetric geometric roles. Our centerline-based parameterization explicitly separates these two factors, improving wall geometry and thickness reconstruction and making it more suitable for BIM-oriented reconstruction.

Figure 7. Qualitative comparison on Structured3D. Each column shows one scene, with rows showing the input point cloud, floor plan baselines, plane baselines, our cuboid reconstruction, and the ground truth. Compared with RoomFormer, CAGE, SceneScript, and SpatialLM, CRAB recovers more coherent wall structures and more accurate door and window geometry in ambiguous, non-Manhattan, and incomplete regions.

Figure 8. Qualitative comparison on real-world scanned scenes. Each row shows one scene: input point cloud (left), SpatialLM (middle), and ours (right). CRAB better preserves wall continuity and opening geometry under practical noise and clutter.

Building Information Modeling

Building Information Modeling (BIM) represents buildings as structured digital models linking geometry, materials, components, and project data.

See all 14 papers

BIMStruct3D: A Fully Automated Hybrid Learning Scan-to-BIM Pipeline with Integrated Topology Refinement
Apr 27, 2026Mahdi Chamseddine, Fabian Kaufmann, Marius Schellen +33D Reconstruction Point Cloud Learning
German Research Center for Artificial Intelligence (DFKI), Kaiserslautern, Germany · RPTU Kaiserslautern-Landau, Kaiserslautern, Germany
BuildAnyPoint: 3D Building Structured Abstraction from Diverse Point Clouds
Feb 27, 2026Tongyan Hua, Haoran Gong, Yuan Liu +33D Mesh Generation Diffusion Models
HKUST(GZ) · Xi’an Jiaotong University · HKUST
Cut a Wall, See a Wall: Identity Substitution in Occluded Point-Cloud Perception for Construction Robots
Oct 25, 2026Tsung-Lung Yang3D Point Cloud Segmentation 3D Semantic Segmentation
National University

Apr 27, 2026J/K move · Enter open · S save

BIMStruct3D: A Fully Automated Hybrid Learning Scan-to-BIM Pipeline with Integrated Topology Refinement

Mahdi Chamseddine, Fabian Kaufmann, Marius Schellen, Christian Glock+2

German Research Center for Artificial Intelligence (DFKI), Kaiserslautern, Germany · RPTU Kaiserslautern-Landau, Kaiserslautern, Germany

3D Reconstruction Point Cloud Learning Building Information Modeling

Open paper AskPDF arXiv

Automatic generation of Building Information Models (BIM) from building scans is a key challenge in architecture and construction. We present a modular pipeline for generating IFC-compliant BIM from 3D point clouds. The hybrid approach combines learning-based semantic segmentation with topology-aware geometric reconstruction to model structural elements accurately. We propose vIoU, adapting voxel-based overlap evaluation to Scan-to-BIM by enabling holistic, instance-matching-free comparison of reconstructed and ground-truth models. We release the German Hospital dataset (DeKH), including high-resolution point clouds, ground truth BIMs, and semantic annotations. Experiments on DeKH and CV4AEC datasets show significant improvements over a RANSAC-based baseline, demonstrating robustness and scalability.

Abstract

Building Information Modeling

Explore similar work

BIMStruct3D: A Fully Automated Hybrid Learning Scan-to-BIM Pipeline with Integrated Topology Refinement

BuildAnyPoint: 3D Building Structured Abstraction from Diverse Point Clouds

Cut a Wall, See a Wall: Identity Substitution in Occluded Point-Cloud Perception for Construction Robots