Rohbau3D

A Shell Construction Site 3D Point Cloud Dataset

Figure: Rohbau3D point cloud feature maps

Abstract

We introduce Rohbau3D, a novel dataset of 3D point clouds that realistically represent indoor construction environments. The dataset comprises 504 high-resolution LiDAR scans captured with a terrestrial laser scanner across 14 distinct construction sites, including residential buildings, a large-scale office complex, educational facilities, and an underground parking garage—all in various stages of shell construction or renovation. Each point cloud is enriched with scalar laser reflectance intensity, RGB color values, and reconstructed surface normal vectors. In addition to the 3D data, the dataset includes high-resolution 2D panoramic renderings of each scene and its associated point cloud features. Designed to reflect the complexity and variability of real-world construction sites, Rohbau3D supports research in geometric processing, scene understanding, and intelligent computing in structural and civil engineering. To our knowledge, it is the first dataset of its kind and scale to be publicly released. Rohbau3D is intended as a foundation for ongoing work, with plans to extend it through additional scenes and targeted annotations to support future research.

Paper

:page_facing_up: Rohbau3D: A Shell Construction Site 3D Point Cloud Dataset

Overview

• Data Records
  • The Scope Of The Data
  • The Dataset Structure
• Installation
• Download and Extract the Data
• Citation
• Acknowledgement

Data Records

The Rohbau3D data records can be summarized as a medium-scale repository of terrestrial laser scan point clouds covering static scenes from a wide variety of shell construction sides. The records include the spatial coordinates annotated with the sensor-specific (1) RGB color, (2) surface reflection intensity information, (3) the reconstruction of surface normal vectors, and (4) panoramic 2D image representations of all feature spaces

The Scope Of The Data

The repository contains in total a set of 504 scenes captured in one of 14 different building environments.

Renovation sites are indicated with an asterisk ().*

The Dataset Structure

rohbau3d
|-- metadata
|   |-- site_list.yaml
|   |-- site_metadata.yaml
|   '-- ... 
|
|-- site_00
|   |-- scan_00000
|   |   |-- coord.npy
|   |   |-- color.npy
|   |   |-- intensity.npy
|   |   |-- normal.npy
|   |   |-- panorama.png
|   |   '-- ...
|   |   
|   |-- scan_00001
|   |-- scan_00002
|   '-- ...
|
|-- site_001
|   |-- scan_01000
|   '-- ...
|
... 
'-- site_013

Installation

Requirements

• git
• pooch
• tqdm
• zstandard
• yaml

Clone Repository

Clone the Rohbau3D Repository to a local space.

git clone https://github.com/RauchLukas/Rohbau3D.git

Conda Environment

Manually create a conda environment and install the package

conda create -n rohbau3d python=3.11 -y
conda activate rohbau3d

cd Rohbau3D
pip install .

Download and Extract the Data

The Dataset can be directly downloaded in chunks from Dataverse | OpenData UniBw M:

Download Link: https://open-data.unibw.de/dataset

Conveniently, this repository offers also the option of downloading the entire dataset or individual pieces using scripts. [RECOMMENDED]

• Short Version:

  Inside the Rohbau3D folder, run the scripts/download.py script to download all dataset point cloud files.

  python scripts/download.py --config config/dataverse.yaml --download --extract
  

  Options:

  • --config [required] : set the path to the configuration script.
  • --download [optional] : Flag to enable download. Default=False.
  • --extract [optional] : Flag to enable file extraction. Default=True.

• Manual Configuration:

  Customize the configuration inside the config/dataverse.yaml file:

  # CONFIGURATION
  # Rohbau3D
  
  # GENERAL
  config_dir: config
  log_dir: log
  log_level: INFO
  
  # DOWNLOAD
  download_hub: dataverse
  download_dir: data/download
  feature_index_file: dataverse_file_index.json
  feature_selection: [all]
  scene_selection: [all]
  
  # FILE EXTRACT
  extract_dir: data/extract
  clean_download_files: False
  

  Options:

  • config_dir : Set the path/to/the/configuration/files location.
  • log_dir: Set the logging path/to/the/logging location.
  • log_level : Set the logging Level.
  • download_hub: Set the download server / hub. [Allowed options: dataverse and default]

    Note: At the moment, the data can only be downloaded from Dataverse https://open-data.unibw.de/.

  • download_dir : Set the path/to/the/download location.
  • feature_index_file : Name the content index file for the download hub.
  • feature_selection : Select the point cloud features to download as a list. Options include:
    • coord : the actual xyz coordinate of the points.
    • color : the RGB color annotation od the points.
    • intensity : the LiDAR reflection intensity annotation of the points.
    • normal : the reconstructed surface normal annotation of the points.
    • all : selects all available point cloud features.
  • extract_dir : Set the path/to/the/file/extraction location.
  • clean_download_files : Set the Flag True, False to delete the download directory at the end of the script.

Citation

If you find our work useful in your research, please cite our paper:

@article{rauch.Rohbau3D.2025,
  title = {Rohbau3D: A Shell Construction Site 3D Point Cloud Dataset},
  shorttitle = {Rohbau3D},
  author = {Rauch, Lukas and Braml, Thomas},
  year = {2025},
  journal = {Scientific Data},
  volume = {12},
  number = {1},
  pages = {1478},
  publisher = {Nature Publishing Group},
  issn = {2052-4463},
  doi = {10.1038/s41597-025-05827-7},
}

Acknowledgement

The surface normal estimation in this repo is based on/inspired by great works, including but not limited to:
SHS-Net