IRIS is a new dataset providing a dense station-based LiDAR point cloud and a CAD model reconstructed close to the points in a semi-automatic way. The scene is a large industrial room (530m²) containing objects of different shapes, colors and sizes such as pipes, valves, pumps and gauges. Due to the complex environment, this dataset is challenging for many computer vision tasks.
IRIS includes a dense point cloud and a CAD model. More details are given on the download page.
We provide 67 LiDAR point clouds from the each of the stations acquisition with high density as well as three clouds merging the stations in three different areas covering in total 30% of the whole scene.
The CAD model is reconstructed close to the points in a semi-automatic way. The tolerance for the reconstruction is 2.56 σ = ±2 cm.
13777859 is the global DOI always refering to the latest dataset version. To cite a specific version, please replace the version number and 13777859 by 13777860 for version 1.0.
@dataset{armangeon_2025_13777859,
author = {Armangeon, Flavien and
Ehret, Thibaud and
Meinhardt-Llopis, Enric and
Grompone Von Gioi, Rafael and
Thibault, Guillaume and
Petit, Marc and
Facciolo, Gabriele},
title = {IRIS: Industrial Room In Saclay},
month = feb,
year = 2025,
publisher = {EDF (Électricité de France)},
version = {1.0},
doi = {10.5281/zenodo.13777859},
url = {https://doi.org/10.5281/zenodo.13777859},
}