"Research" on Feature Representation Method of Deep Learning for 3D CAD Model Clustering

The traditional feature extraction and representation methods for 3D CAD model clustering are now in the bottleneck, and the complex process leads to the increasing of losses so that the clustering results is not satisfactory . Therefore, A method based on deep learning is proposed to learn and represent the features of 3D CAD model. Convolutional neural network and Autoencoder are used to represent and learn the adjacency matrix of attributes of 3D CAD model, and the low-dimensional eigenvectors of the hidden layer are used as the only representation of the model. From the final clustering results, the NMI and V-measure values obtained by this method are improved compared with the traditional methods. The clustering results can effectively support the reuse of the model by engineers.