Know3D: Prompting 3D Generation with Knowledge from Vision-Language Models

TL;DR

Know3D introduces a framework that leverages multimodal large language models to incorporate semantic knowledge into 3D generation, enabling controllable and semantically guided reconstruction of unobserved regions.

Contribution

The paper presents a novel method that injects knowledge from vision-language models into 3D generative models, improving control and semantic consistency in unobserved regions.

Findings

Semantic guidance improves back-view generation control.

Knowledge injection reduces stochasticity in 3D hallucination.

Framework bridges textual instructions and 3D geometric reconstruction.

Abstract

Recent advances in 3D generation have improved the fidelity and geometric details of synthesized 3D assets. However, due to the inherent ambiguity of single-view observations and the lack of robust global structural priors caused by limited 3D training data, the unseen regions generated by existing models are often stochastic and difficult to control, which may sometimes fail to align with user intentions or produce implausible geometries. In this paper, we propose Know3D, a novel framework that incorporates rich knowledge from multimodal large language models into 3D generative processes via latent hidden-state injection, enabling language-controllable generation of the back-view for 3D assets. We utilize a VLM-diffusion-based model, where the VLM is responsible for semantic understanding and guidance. The diffusion model acts as a bridge that transfers semantic knowledge from the VLM to the 3D generation model. In this way, we successfully bridge the gap between abstract textual instructions and the geometric reconstruction of unobserved regions, transforming the traditionally stochastic back-view hallucination into a semantically controllable process, demonstrating a promising direction for future 3D generation models.