Supplementary material for "Augmented Carpentry: a Computer Vision-based Assistance for Manual Saw-cutting and Drilling in Timber Construction"

Dataset content

The present data set has been used to evaluate timber artifacts produced with AR-assisted fabrication methods.

The raw scans of the produced mock-ups, as well as their execution model, can be inspected in the dataset in the folder raw_data/. Two options are possible to access the dataset since we made all the data available in multiple interoperable formats:

1. via Rhinoceros 3D: all the point cloud and execution model files with the extension .3dm
2. other visualizer/engines or CAD environments: for all point cloud files .ply or .e57

Each entry of the raw_data/ dataset folder is structured as follow:

• <name_of_the_mockup>
  • a_beams: contains a list of scans of each individual beam composing the mock-up
  • b_raw_scan: contains the raw scan of the entire structure
  • c_rh_overview: contains the 3D execution model as well as all the previous point clouds

Reproducible analysis pipeline

The basic analysis of the point cloud-CAD comparison is done via diffCheck: a Grasshopper Plug-in that allows to identify discrepancies across point clouds and 3D models of both individually machined timber pieces featuring various joints as well as fully assembled timber structures. It can help you quantify the differences between the CAD and scanned fabricated structures, providing a comprehensive report highlighting the discrepancies. These reports are present in the results/ folder:

• <name_of_the_mockup>:
  • a_beams:
    • <beam_id>:
      • <beam_id>_joint.csv: values of the joint location analysis from diffCheck
      • <beam_id>_jointfaces.csv: values of the joint faces quality analysis from diffCheck
      • <beam_id>_joint_<timestamp>.diffCheck: native export for joint location analysis from diffCheck
      • <beam_id>_jointfaces_<timestamp>.diffCheck: native export for joint faces quality analysis from diffCheck
      • <beam_id>_trm_cloud.ply: the compared raw point cloud registered to the corresponding CAD model
  • b_assembly:
    • <name_of_the_mockup>_beam.csv: values of the assembly analysis from diffCheck
    • <name_of_the_mockup>_<timestamp>.diffCheck: native format of the assembly analysis from diffCheck
    • <name_of_the_mockup>_legend.png: color legend of the analysis
    • <name_of_the_mockup>_trm_cloud.ply: transformed raw point cloud to the analysed 3D model
    • <name_of_the_mockup>_trm_cloud_clr.ply: transformed raw point cloud to the analysed 3D model with color indicating the measured error

for all the analysed structures both:

• as .diffCheck files, and they can be read only via Rhinoceros 3D via the dedicated import component
• as .csv files that can be parsed and read in any environment

In addition, the statistical evaluation presented in section 3 of the paper can be verified with a Python script:

• main.py: the file containing the analysis
• envirnoment.eval.yml: this is the configuration file to run to create the conda environment to run the analysis within
• csv files: from the diffCheck analysis in the results/ folder

For the usage of the Python script:

e.g.:

Organization

Ecole Polytechnique Fédérale de Lausanne (EPFL), School of Architecture, Civil and Environmental Engineering(ENAC), Institute of Civil Engineering (IIC), Laboratory for Timber Constructions (IBOIS), GC H2 711, Station18, Lausanne, 1015, Vaud, Switzerland. (Contacts: andrea.settimi@epfl.ch)