Spectral Mesh Comparison Methods for Generative Design
Recently, generative methods have been utilized as mass-customization tools with 3D printing being a means for production to give users the opportunity to generate unique and customizable objects. As generative methods allow the creation of complex geometries by automation and customization by parameterization, 3D printing has become the only viable solution that enables the transformation of arbitrary generated output into physical form, without losing the benefits of the generative method. Here, we leverage spectral methods for shape similarity and dissimilarity measurements based on the discrete laplace-beltrami operator, such as ShapeDNA, Global point signature or Weighted Spectral Distances to evaluate and define the expressivity of generative methods. These metrics allow a user or the generative system to assess the uniqueness of the generated objects themselves within the space of the generated objects. Thereby, we offer a strategy for the evaluation of generative methods and their associated artifacts for improved mass-customization methods.