OBJECT RELATIONSHIP ESTIMATION FROM A 3D SEMANTIC MESH

摘要

Implementations disclosed herein provide systems and methods for determining relationships between objects based on the original semantic mesh of vertices and faces representing the 3D geometry of a physical environment. Such original semantic meshes can be created and used to provide input to machine learning models that estimate relationships between objects within the physical environment. For example, the machine learning model may output a graph of nodes and edges indicating that the vase is on top of a table, or that a specific instance V1 of the vase is on top of a specific instance T1 of the table.