[论文] ELSA3D: Elastic Semantic Anchoring for Unified 3D Understanding and Ge...

论文概要

研究领域: CV 作者: Tianjiao Yu, Xinzhuo Li, Yifan Shen 发布时间: 2025-07-09 arXiv: 2507.06842

中文摘要

统一的3D基础模型希望在单一骨干网络中生成3D资产并用语言推理，但其文本-3D交互仍然主要是隐式的。现有方法将文本和3D令牌连接成扁平序列并依赖自注意力，将粗粒度结构线索和细粒度几何细节混为一种无差别的表示。我们引入ELSA3D，一种通过弹性语义锚定来解决这个问题的统一3D模型，沿匹配抽象尺度共同构建语言和几何推理。ELSA3D使用尺度感知八叉树分词器表示几何，并引入锚定令牌——稀疏的跨模态单元，选择语义线索、将其路由到最相关的3D尺度、检索尺度特定的几何证据，并将融合信号写回统一表示中，保持交互稀疏而精确。轻量级的逐块路由器使计算和推理都具有弹性，选择哪些文本令牌在哪些几何尺度上实例化锚定，使跨模态能力集中在最需要对齐的地方。ELSA3D在图像到3D生成、文本到3D生成和3D描述方面实现了最先进的性能，超越了最强的统一基线，同时相对于同一模型的非弹性版本，FLOPs和推理延迟大致减半。

原文摘要

Unified 3D foundation models aspire to generate 3D assets and reason about them in language within a single backbone, but their text-3D interaction remains largely implicit. Existing methods concatenate text and 3D tokens into a flat sequence and rely on self-attention, collapsing coarse structural cues and fine geometric details into one undifferentiated representation. We introduce ELSA3D, a unified 3D model that addresses this with elastic semantic anchoring, structuring language and geometric reasoning jointly along matched abstraction scales. ELSA3D represents geometry with a scale-aware octree tokenizer and introduces Anchor Tokens, sparse cross-modal units that select semantic cues, route them to the most relevant 3D scale, retrieve scale-specific geometric evidence, and write the f...

--- *自动采集于 2026-07-09*

#论文 #arXiv #CV #小凯