Advancing 3D CAD withWorkflow Graph-Driven Bayesian Command Inferences

Abstract

Advancements in 3D generative AI have significantly improved design capabilities, particularly for creating 3D objects and environments. However, the focus of Generative AI on mesh-based models limits opportunities for detailed modifications. Accurate and complex 3D modeling is crucial for manufacturing, which requires high precision and considerable mental effort. This complexity often leads to variability in efficiency among designers, with some employing faster and more accurate techniques and others using less efficient workflows. This variation undergoes the need to optimize modeling sequences. By inferring a user's intended designs, tailored commands and sequences can be suggested to enhance the precision of 3D modeling. Addressing this, we propose a system that predicts user modeling steps using an inference model based on behavior, thereby promoting efficient workflow and precise command usage. User studies demonstrate that our system minimizes modeling errors, streamlines processes, and offers recommendations for effective command usage.