HybridCAD++: Expanded Dataset for Hybrid Additive-Subtractive Manufacturing Feature Recognition in B-Rep CAD Models

<p>The<strong> <em>HybridCAD++</em> </strong>dataset is a significantly expanded version of the <strong><em>HybridCAD</em></strong> dataset, offering a larger volume of CAD models and a more comprehensive set of hybrid additive-subtractive manufacturing features. This dataset includes additional feature classes, bringing the total to <strong>36</strong>, and contains over <strong>161,000 samples</strong>&mdash;making it a unique and robust resource for machine learning applications in hybrid manufacturing feature recognition.</p> <h3>Key Differences from HybridCAD</h3> <ul> <li><strong>Increased Dataset Volume</strong>: <em>HybridCAD++</em> features a total of 161,441 CAD models, significantly larger than the original <em>HybridCAD</em> dataset.</li> <li><strong>Expanded Feature Classes</strong>: This dataset includes 36 feature labels, with newly added classes. This increase in feature variety enhances the dataset's applicability to complex hybrid manufacturing scenarios.</li> </ul> <h3>Dataset Composition</h3> <p>The dataset includes three primary components:</p> <ol> <li> <p><strong>STEP Files</strong>:</p> <ul> <li>Each CAD model is stored in STEP format and includes labeled B-Rep faces for hybrid manufacturing feature recognition.</li> <li>The CAD models were generated programmatically using PythonOCC, ensuring consistent quality and scalability.</li> </ul> </li> <li> <p><strong>Feature Labels</strong>:</p> <ul> <li><strong>File</strong>: <code>feature_labels.txt</code></li> <li>Contains label IDs for each hybrid additive-subtractive feature across B-Rep faces in each CAD model.</li> <li>With 36 unique feature classes, this file allows precise mapping of CAD model faces to specific hybrid features.</li> </ul> <ul> <li>&nbsp;</li> </ul> </li> <li> <p><strong>Hierarchical B-Rep Graphs</strong>:</p> <ul> <li>Stored in HDF5 format, these graphs provide structured access to B-Rep data, as detailed in <code>h5_structure.txt</code>.</li> </ul> </li> </ol> <h3>Dataset Splits</h3> <p>The dataset is divided into three subsets, structured for effective model training and evaluation:</p> <ul> <li><strong>Training Set</strong>: 113,008 samples (70%)</li> <li><strong>Validation Set</strong>: 32,288 samples (20%)</li> <li><strong>Testing Set</strong>: 16,145 samples (10%)</li> </ul> <h3>Full Feature Label List</h3> <p>This comprehensive list includes both subtractive and additive manufacturing features, with added classes for more intricate hybrid manufacturing applications:</p> <p>&nbsp;</p> <p>Label &nbsp; &nbsp;Feature<br>0 &nbsp; &nbsp;Chamfer<br>1 &nbsp; &nbsp;Through hole<br>2 &nbsp; &nbsp;Triangular passage<br>3 &nbsp; &nbsp;Rectangular passage<br>4 &nbsp; &nbsp;6-sided passage<br>5 &nbsp; &nbsp;Triangular through slot<br>6 &nbsp; &nbsp;Rectangular through slot<br>7 &nbsp; &nbsp;Circular through slot<br>8 &nbsp; &nbsp;Rectangular through step<br>9 &nbsp; &nbsp;2-sided through step<br>10 &nbsp; &nbsp;Slanted through step<br>11 &nbsp; &nbsp;O-ring<br>12 &nbsp; &nbsp;Blind hole<br>13 &nbsp; &nbsp;Triangular pocket<br>14 &nbsp; &nbsp;Rectangular pocket<br>15 &nbsp; &nbsp;6-sided pocket<br>16 &nbsp; &nbsp;Circular end pocket<br>17 &nbsp; &nbsp;Rectangular blind slot<br>18 &nbsp; &nbsp;Vertical circular end blind slot<br>19 &nbsp; &nbsp;Horizontal circular end blind slot<br>20 &nbsp; &nbsp;Triangular blind step<br>21 &nbsp; &nbsp;Circular blind step<br>22 &nbsp; &nbsp;Rectangular blind step<br>23 &nbsp; &nbsp;Round<br>24 &nbsp; &nbsp;Extrude cylinder<br>25 &nbsp; &nbsp;Extrude rectangle<br>26 &nbsp; &nbsp;Extrude triangle<br>27 &nbsp; &nbsp; Extrude hexagon<br>28 &nbsp; &nbsp; Extrude pentagon<br>29 &nbsp; &nbsp; &nbsp;Elliptical/Oval blind hole<br>30 &nbsp; &nbsp; &nbsp;Elliptical/Oval through hole<br>31 &nbsp; &nbsp; &nbsp;Slot hole<br>32 &nbsp; &nbsp; &nbsp;Obround boss<br>33 &nbsp; &nbsp; &nbsp;5-sided passage<br>34 &nbsp; &nbsp; &nbsp;5-sided pocket<br>35 &nbsp; &nbsp; &nbsp;Cylinder with hole<br>36 &nbsp; &nbsp; &nbsp;Stock</p>