3D 视觉与重建 · CVPR2026 ·2026.06.03

Vinedresser3D:迈向智能体式文本引导 3D 编辑

Vinedresser3D: Towards Agentic Text-guided 3D Editing

Yankuan Chi、Xiang Li、Zixuan Huang、James Matthew Rehg

与你有关的部分

把 3D 编辑搬到原生 3D 生成模型的潜空间里做,用 MLLM 做编辑意图解析与区域定位、再用 inversion-based rectified-flow 图像修复提供视觉引导,这条「语言代理 + 2D 引导 + 3D 潜空间修复」的流水线可直接迁移到 3D 资产编辑与可控 3D 生成任务。

原文依据 an inversion-based rectified-flow inpainting pipeline with an interleaved sampling module performs editing in the 3D latent space
可迁移到 文本引导 3D 编辑3D 资产生成图像修复多模态指令理解

摘要译文

由模型翻译 · 术语统一 · 请对照原文核对关键结论

文本引导的 3D 编辑旨在使用自然语言指令修改已有的 3D 资产。现有方法难以同时理解复杂提示词、在 3D 中自动定位编辑区域并保持未编辑内容不变。我们提出 Vinedresser3D,一个面向高质量文本引导 3D 编辑的智能体框架,它直接在原生 3D 生成模型的潜空间中操作。给定一个 3D 资产和一条编辑提示词,Vinedresser3D 使用多模态大语言模型推断原始资产的丰富描述,识别需要编辑的部分与编辑类型(添加、修改、删除),并生成分解到结构与外观层面的文本引导。随后,该智能体选择信息量丰富的视角,并应用图像编辑模型以获得视觉引导。最后,一条基于反演的整流流图像修复流水线配合交错采样模块,在 3D 潜空间中执行编辑,在保持 3D 一致性与未编辑区域的同时强化提示词对齐。在多种 3D 编辑任务上的实验表明,Vinedresser3D 在自动指标与人类偏好研究中均优于已有基线,并实现了精确、连贯且无需掩码的 3D 编辑。

英文原文摘要 1216 字符

Text-guided 3D editing aims to modify existing 3D assets using natural-language instructions. Current methods struggle to jointly understand complex prompts, automatically localize edits in 3D, and preserve unedited content. We introduce Vinedresser3D, an agentic framework for high-quality text-guided 3D editing that operates directly in the latent space of a native 3D generative model. Given a 3D asset and an editing prompt, Vinedresser3D uses a multimodal large language model to infer rich descriptions of the original asset, identify the parts to edit and the edit type (addition, modification, deletion), and generate decomposed structural and appearance-level text guidance. The agent then selects an informative view and applies an image editing model to obtain visual guidance. Finally, an inversion-based rectified-flow inpainting pipeline with an interleaved sampling module performs editing in the 3D latent space, enforcing prompt alignment while maintaining 3D coherence and unedited regions. Experiments on diverse 3D edits demonstrate that Vinedresser3D outperforms prior baselines in both automatic metrics and human preference studies, while enabling precise, coherent, and mask-free 3D editing.

标签

会议 CVPR2026
页码 12673-12683
Bibkey
来源 CVF Open Access