Dataset for the paper "A framework for robotic excavation and dry stone construction using on-site materials"

<p>Stone data from the <i>Science Robotics</i> paper "A framework for robotic excavation and dry stone construction using on-site materials" containing:</p><ul><li>Mesh files of 1,100 stones (quarried boulders, erratics, and concrete debris) that were digitized by the autonomous excavator HEAP<ul><li><a href="https://zenodo.org/api/records/10038881/draft/files/1100%20Unprocessed%20Stone%20Meshes.zip/content">1100 Unprocessed Stone Meshes.zip: </a>Raw mesh files directly from the poisson reconstruction of accumulated LiDAR points, containing some artifacts and floating geometries</li><li><a href="https://zenodo.org/api/records/10038881/draft/files/1100%20Closed%20Stone%20Meshes.zip/content">1100 Closed Stone Meshes.zip: </a>Clean, closed, downsampled meshes</li><li><a href="https://zenodo.org/api/records/10038881/draft/files/Stone_Shape_Properties.csv/content">Stone_Shape_Properties.csv: </a>Properties file with a list of the stone IDs (IDs in the 1xxx and 3xxx range typically correspond to concrete elements) and select shape properties</li></ul></li><li>A dataset of candidate placements from automatically generated stone walls. &nbsp;The candidate placement data zip files contain:<ul><li><a href="https://zenodo.org/api/records/10038881/draft/files/Candidate_Placement_Data-npy.zip/content">Candidate_Placement_Data-npy.zip: </a>SDF (.npy) representation of each candidate, with three channels of 32x32x32 for distances to the stone, the already-placed stones, and the target wall</li><li><a href="https://zenodo.org/api/records/10038881/draft/files/Candidate_Placement_Data-pcd.zip/content">Candidate_Placement_Data-pcd.zip: </a>Point cloud (.pcd) representations of each candidate, with separate files for the placed stone, target wall (search volume), and already-placed stones (where they exist)</li></ul></li><li><a href="https://zenodo.org/api/records/10038881/draft/files/sdf_classifier.zip/content">sdf_classifier.zip: </a>Python examples:<ul><li>Rendering the three channel SDF data to mesh geometry using marching cubes and libigl</li><li>Candidate SDF classification using the pretrained model</li></ul></li><li>Candidate attributes and labels<ul><li><a href="https://zenodo.org/api/records/10038881/draft/files/candidate_attributes_labels.csv/content">candidate_attributes_labels.csv: </a>CSV file containing a list of UUID's corresponding to each candidate placement in the dataset. &nbsp;For each candidate, additional information is included about the dimensions and location of the solution, together with the (subjectively) hand-labelled binary value for placement viability.&nbsp;</li></ul></li><li><a href="https://zenodo.org/api/records/10038881/draft/files/README.md/content">README.md: </a>An additional readme with some details on the&nbsp;attributes file</li></ul><p>If you use this data in your research, please cite the <a href="https://www.science.org/doi/10.1126/scirobotics.abp9758">journal article</a>.</p><p>&nbsp;</p>