Chorus：面向整体 3D 高斯场景编码的多教师预训练
Chorus: Multi-Teacher Pretraining for Holistic 3D Gaussian Scene Encoding
把 2D 基础模型（语言对齐、通用、物体感知教师）的互补信号蒸馏进一个共享的前馈 3DGS 场景编码器，得到可直接用于开放词表分割、线性探测和 LLM 问答的通用 3D 高斯特征；其 render-and-distill 适配与仅用高斯中心/颜色/法向的变体，为点云等非 3DGS 表示提供了可复用的预训练迁移路径。
原文依据 a multi-teacher pretraining framework that learns a holistic feed-forward 3D Gaussian Splatting (3DGS) scene encoder by distilling complementary signals from 2D foundation models
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论尽管 3DGS 已成为一种高保真场景表示，但直接从其基元中编码丰富、通用的特征仍鲜有探索。为填补这一空白，我们提出 Chorus，一个多教师预训练框架，通过从 2D 基础模型中蒸馏互补信号，学习整体的前馈 3D 高斯泼溅（3DGS）场景编码器。Chorus 采用共享的 3D 编码器与教师专属投影器，从语言对齐、通用以及物体感知的教师中学习，促使形成一个共享嵌入空间，以捕获从高层语义到细粒度结构的信号。我们在广泛的任务上评估 Chorus：开放词表语义分割与实例分割、线性探测与解码器探测、数据高效监督，以及基于 LLM 的问答。除 3DGS 之外，我们还仅使用高斯的中心、颜色与估计法向预训练了一个变体，并在若干仅支持点云的基准上测试 Chorus。有趣的是，该编码器展现出很强的迁移能力，在使用少 39.9 倍的训练场景的情况下优于点云基线。最后，我们提出一种渲染-蒸馏适配方法，便于进行域外微调。我们的代码与模型已发布于该代码库。
英文原文摘要 1254 字符
While 3DGS has emerged as a high-fidelity scene representation, encoding rich, general-purpose features directly from its primitives remains under-explored. We address this gap by introducing Chorus, a multi-teacher pretraining framework that learns a holistic feed-forward 3D Gaussian Splatting (3DGS) scene encoder by distilling complementary signals from 2D foundation models. Chorus employs a shared 3D encoder and teacher-specific projectors to learn from language-aligned, generalist, and object-aware teachers, encouraging a shared embedding space that captures signals from high-level semantics to fine-grained structure. We evaluate Chorus on a wide range of tasks: open-vocabulary semantic and instance segmentation, linear and decoder probing, data-efficient supervision, as well as LLM-based Q&A. Besides 3DGS, we also test Chorus on several benchmarks that only support point clouds by pretraining a variant using only Gaussians' centers, colors, estimated normals. Interestingly, this encoder shows strong transfer and outperforms the point clouds baseline while using 39.9xfewer training scenes. Finally, we propose a render-and-distill adaptation that facilitates out-of-domain finetuning. Our code and model is released at this codebase.