Subset of ShapeNet-Part 3D Point Cloud Dataset for Symmetry-Aware Deep Learning for SOAP

This dataset contains 3D point cloud data derived from the ShapeNet-Part dataset, prepared for use with the Smooth Overlap of Atomic Positions (SOAP) descriptor. The data are intended for experiments exploring SOAP or other .xyz file friendly preprocessing method for 3D point cloud analysis. The dataset includes both .txt and .xyz formats. The .txt files contain the subset of the original point cloud data with labels, while the .xyz files were generated from the .txt data and are provided for convenience in workflows that require the XYZ format. Each file represents an airplane along with its corresponding class labels. Dataset structure Test/ TXT/: Labeled test point clouds in .txt format XYZ/: Labeled test point clouds in .xyz format Subdirectories represent different training set sizes, for example:TXT0.1, TXT1, TXT10, TXT100, XYZ0.1, XYZ1, XYZ10, XYZ100, etc. Each folder contains the corresponding portion of the training dataset (e.g., 0.1 = 0.1% of the full training set, 100 = full dataset). File format .txt: ASCII-encoded point clouds with labels .xyz: XYZ coordinate files with associated labels (_labeled.xyz and _labeled.label)