An Assembly Support System Based on 3D Part Retrieval via Natural Language

Abstract

We propose an approach that leverages existing high-quality 3D model datasets by retrieving and combining components at the part level. Specifically, the user inputs a textual description of a desired part, such as “chair leg,” and the system retrieves multiple candidate parts from a database, from which the user selects and places a preferred one by clicking. By repeating this process, the system aims to enable even users without specialized knowledge to incrementally assemble a 3D model.