Materials Follow Form and Function: Probabilistic Factor Graph Approach for Automatic Material Assignments to 3D Objects
There are strong co-relations between material assignment, shape, and functionality of a part in overall product/assembly. However, these strong co-relations are rarely exploited for automated material assignment. We present a probabilistic graphical model-based approach to automatically assign materials to the parts (components) of a 3D object (assembly). The presented model performs material assignment by identifying the relations between shape, functionality, and materials of parts in the existing database objects. By learning the context dependent correlation without supervision from a set of objects and their segmented parts, the learned model can be used to assign proper real materials to the parts of a query object. Our primary contributions are: a) the real materials definition and assignment and b) assigning materials based on the functionality and form of the parts in the object. The performance of proposed computational approach is demonstrated by results of material assignment on various query objects without pre-specified material definitions.