Scans dataset for evaluation of AR-based assembly pipeline of half-timber dry-stone structures

Dataset description

The dataset is realized by the EPFL Laboratories of IBOIS, and EESD on the occasion of joint collaboration with the common goal of exploring augmented reality (AR)-based fabrication for half-timber dry-stone construction. It is open sourced and free to be used for anyone's own research.

The current dataset has been employed to evaluate the designed digital fabrication pipeline. It consists of raw point clouds and reconstructed models of two digitized prototypes of one-layered dry-stone structures with a 1.7 m length, ~0.6 height, and 0.7 m width, each with a different composition of building units. The first wall presents 40 mineral by-products uniquely from by-products of quarry processes involving sawing and water-cutting. In contrast, the second wall features 30 mineral scraps issued from all mixed quarry's operations of transformations.

The collection contains both the reconstructed models of the as-built artifacts and the corresponding recorded model of the AR-guided assembly processes for the two walls. The recorded model contains meshes of the placed stones accordingly to the developed geometric planner. By comparing the misalignment between each stone of both models (e.g. Root Mean Square Error distances between two set of points) is possible to obtain metrics about the performance of the proposed AR-assembly method.

Dataset labeling guide

label
content

01/02
First or second wall.

rs
Raw data: unprocessed scans.

ab
As-built model: model presenting all the registered stones.

rc
Recorded model: collection of tracked stones during the construction. Each stone has being placed following the AR system is recorded in this sub-set.

ev
Evaluation set: documentation and working files for the evaluation of the developed AR assembly system.

colored_stone
Evaluation set outcome: a graphical representation of the misalignment between each stone of the as-built model (ab) with the recorded model (rc).

Equipement specs

All the raw scans present in the dataset have been obtained from a FARO Freestyle 2 equipped with a Mobile PC for live point cloud processing.

@manual{farofreestyle2,
title = {FARO Freestyle 2 and Mobile PC},
year = 2022,
month = feb,
note = {User Manual},
organization = {FARO Technologies Inc.},
url = {https://downloads.faro.com/index.php/s/sqcRBipgSy9GaEq?dir=undefined&openfile=138985}
}

For extra info

The open-sourced code for the developed AR assembly:

https://doi.org/10.5281/zenodo.7181087

The complete dataset of digitized mineral scraps:

https://doi.org/10.5281/zenodo.7189478

Version notes

- The current dataset is published as linked documentation to a future publication, yet not reviewed.

Change log

- Typos

- Correct order of authors