Rohbau3D-Semantics: Dense Semantic and Instance Labels for Point Clouds of Indoor Shell Construction Scenes
Organizations: University of the Bundeswehr Munich, Germany
Abstract
We present the Rohbau3D-Semantics dataset, which adds dense semantic and instance-level labels to laser scans of the previously released Rohbau3D domain data from shell construction sites. The dataset comprises 504 point cloud scenes from 14 survey sites, each captured with a terrestrial laser scanner and stored in a scanner-centered local coordinate system. Unlike point clouds sampled from reconstructed meshes or synthetic building models, the data preserve the characteristic properties of real laser scanning, including non-uniform sampling, measurement artifacts, reflections, and fine geometric detail. Combined with manual labels, expert oversight during pointwise annotation, and full-scene review, this yields a high-quality benchmark for semantic scene understanding in complex built environments. The dataset includes 17 class labels of building components and one background class, along with instance annotations for object-level analysis. In addition to the 3D point clouds, the release also provides scripts for deriving geometrically aligned panoramic and cube-map representations of color, LiDAR reflection intensity, depth, surface normals, semantic labels, and instance labels from the 3D arrays. We describe the acquisition pipeline, annotation protocol, quality-control process, and panorama-generation procedure, and provide benchmark splits and baseline experiments for semantic segmentation. By combining precise terrestrial laser scans, dense annotations, and a wide variety of shell-construction scenes, Rohbau3D-Semantics bridges the gap among synthetic indoor datasets, mesh-derived benchmarks, and smaller construction-specific resources. The dataset is intended to support research on 3D scene understanding, construction-site analytics, and multimodal scene analysis in the built environment. Access and download the full dataset via https://github.com/RauchLukas/rohbau3d
Figures & tables
| Dataset | Domain | Extent | Reported scale | Sensors | Classes | Sem./Inst. | Modalities | R/S |
|---|---|---|---|---|---|---|---|---|
| S3DIS [ 2 ] | Interior | 6 areas / 271 rooms | m 2 / 215M pts | RGB-D | 13 | Y/Y | P, RGB, D, N | R |
| ScanNet v2 [ 4 ] | Interior | 1,513 scans | 2.5M frames | RGB-D | 20 | Y/Y | M, RGB, D | R |
| ScanNet++ v2 [ 21 ] | Interior | 1006 scenes | 1.1M DSLR+10M iPhone frames | TLS+DSLR+RGB-D | 3.4k | Y/Y | P, M, RGB, D | R |
| LiDAR-Net [ 6 ] | Interior | 14 areas / 465 rooms | m 2 / 3.6B points | Mobile LiDAR | 24 | Y/Y | P, RGB | R |
| VASAD [ 9 ] | BIM | 6 models | 62,000 m 2 | Sim. LiDAR | 11 | Y/N | P, M | S |
| Hilti-Oxford [ 31 ] | Constr. | 2 sites / 16 sequences | 3,000 m 2 | LiDAR+cam. | – | N/N | P, RGB, IMU | R |
| Site | Points | Instances | mean Pts./scene | mean Inst./scene |
|---|---|---|---|---|
| site_00 | 629,986,448 | 1,227 | 17,499,620 | 34.08 |
| site_01 | 224,578,587 | 562 | 10,694,220 | 26.76 |
| site_02 | 118,823,353 | 622 | 10,802,120 | 56.55 |
| site_03 | 1,549,189,153 | 3,366 | 27,664,090 | 60.11 |
| site_04 | 378,528,127 | 998 | 27,037,720 | 71.29 |
| site_05 | 1,771,487,032 | 2,134 | 27,253,650 | 32.83 |
| model | None | Ceiling | Slanted Ceiling | Ceiling Cutout | Floor | Wall | Drywall | Wall Cutout | Parapet | Beam | Column | Staircase | Window Cutout | Window | Door Rough Open. | Facade Element | Elevator Doors | TGA | mIoU |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| PointNet++ | 0.58 | 0.95 | 0.45 | 0.42 | 0.95 | 0.71 | 0.28 | 0.19 | 0.30 | 0.48 | 0.49 | 0.69 | 0.11 | 0.13 | 0.32 | 0.18 | 0.14 | 0.57 | 0.44 |
| PointNeXt-S | 0.61 | 0.95 | 0.48 | 0.32 | 0.95 | 0.76 | 0.46 | 0.11 | 0.44 | 0.47 | 0.52 | 0.64 | 0.06 | 0.06 | 0.29 | 0.23 | 0.02 | 0.57 | 0.44 |
| PointNeXt-L | 0.67 | 0.96 | 0.58 | 0.63 | 0.97 | 0.85 | 0.65 | 0.23 | 0.57 | 0.59 | 0.73 | 0.81 | 0.22 | 0.12 | 0.50 | 0.36 | 0.03 | 0.62 | 0.56 |
| Swin3D-S | 0.75 | 0.98 | 0.73 | 0.83 | 0.97 | 0.87 | 0.68 | 0.37 | 0.76 | 0.64 | 0.81 | 0.93 | 0.43 | 0.32 | 0.60 | 0.59 | 0.45 | 0.69 | 0.69 |
| Swin3D-L | 0.77 | 0.97 | 0.72 | 0.75 | 0.97 | 0.91 | 0.78 | 0.30 | 0.68 | 0.69 | 0.78 | 0.93 | 0.45 | 0.56 | 0.63 | 0.59 | 0.46 | 0.66 | 0.70 |
| PTv3 | 0.77 | 0.98 | 0.73 | 0.80 | 0.98 | 0.89 | 0.72 | 0.46 | 0.62 | 0.67 | 0.82 | 0.92 | 0.48 | 0.43 | 0.63 | 0.64 | 0.71 | 0.68 | 0.72 |
Appendix figures & tables4 assets
Supplementary material from the paper’s appendix.
Appendix
| Site | Building and construction | Distinctive geometry and sensing conditions |
|---|---|---|
| site_00 | Multi-story apartment building; brick masonry | Small–medium rooms; partly plastered walls; windows installed, doors absent. |
| site_01 | Multi-story apartment building; brick masonry | Small–medium rooms; slanted ceilings; partly plastered walls; windows installed, doors absent. |
| site_02 | Underground parking structure; reinforced concrete | Column grid; variable ceiling heights; low light; floor puddles. |
| site_03 | Multi-story school; reinforced-concrete frame | Large rooms; well-lit areas; floor puddles. |
| site_04 | Indoor swimming-pool hall; reinforced concrete | Large open volume; rounded ceiling components; large floor opening; facade not installed. |
| site_05 | Multi-story school; drywall partitions | Variable room sizes; semi-transparent temporary facade covering. |
Point Cloud Understanding
Point cloud understanding turns sparse 3D measurements into object, scene, shape, and motion information for machines.
See all 48 papers