[论文] A Unified Framework for Image-to-3D Part Generation via Variable Granu...

论文概要

研究领域: CV 作者: Qiwen Gu, Xin Zhou, Xiwu Chen, Zhongyang Zhu, Zhuo Zhang, Junqiao Zhao 发布时间: 2026-10-22 arXiv: 2610.31671

中文摘要

可控的 3D 部件生成对数字资产创作至关重要，但现有方法缺乏灵活性，难以适应不同用户工作流中不同程度的控制需求。本文提出 FlexPart——一个统一的图像到 3D 部件生成框架，在单一模型内支持可变粒度的几何条件输入。通过门控自适应调制机制，FlexPart 无缝整合了从稀疏点到稠密边界框和掩码的异构输入，释放出强大的跨模态协同能力。我们还提出了一种非对称几何引导机制，利用强空间先验对不同粒度的表示进行对齐，从而改善几何一致性并促进任务间协同。在 PartVerseXL 数据集的掩码条件下，FlexPart 将 Part CD 相比 TripoSG + SAM3D 降低了 16.6%，相比 OmniPart 降低了 10.5%，同时保持了有竞争力的整体网格质量。

原文摘要

Controllable 3D part generation is pivotal for digital asset creation, yet existing methods lack the flexibility to accommodate diverse user workflows requiring varying levels of control. In this paper, we introduce FlexPart, a unified image-to-3D part generation framework that supports variable-granularity geometric conditions within a single model. By seamlessly integrating heterogeneous inputs, ranging from sparse points to dense bounding boxes and masks, via gated adaptive modulation, FlexPart unlocks a powerful cross-modal synergy. We also propose an asymmetric geometric guidance mechanism that leverages strong spatial priors to align representations across granularities, improving geometric consistency and facilitating inter-task synergy. Under mask conditioning on PartVerseXL, FlexP...

#论文 #arXiv #CV #小凯