LightSplat：五秒内完成快速且内存高效的开放词表三维场景理解
LightSplat: Fast and Memory-Efficient Open-Vocabulary 3D Scene Understanding in Five Seconds
把语义压缩成 2 字节索引、只挂在显著区域并用索引-特征映射管理，这套免训练、免逐高斯特征优化的做法可直接迁移到任何基于 3DGS 的开放词表查询或编辑管线，把原本分钟级的特征蒸馏降到秒级。
原文依据 injects compact 2-byte semantic indices into 3D representations from multi-view images
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论开放词表三维场景理解使用户能够通过自然语言在复杂三维环境中分割新物体。然而，现有方法由于迭代优化和逐高斯稠密特征分配而仍然速度慢、内存开销大且过于复杂。为此，我们提出 LightSplat，一个快速且内存高效的免训练框架，它从多视图图像向三维表示中注入紧凑的 2 字节语义索引。通过仅向显著区域分配语义索引，并用轻量级索引-特征映射对其进行管理，LightSplat 消除了昂贵的特征优化与存储开销。我们进一步通过单步聚类在三维中关联几何与语义相关的掩码，从而保证语义一致性并实现高效推理。我们在 LERF-OVS、ScanNet 和 DL3DV-OVS 上跨复杂室内外场景评估了本方法。结果表明，LightSplat 取得了最优性能，加速最高达 50-400 倍，内存降低 64 倍，使可扩展的语言驱动三维理解成为可能。更多细节请访问项目主页 https://vision3d-lab.github.io/lightsplat/。
英文原文摘要 1152 字符
Open-vocabulary 3D scene understanding enables users to segment novel objects in complex 3D environments through natural language. However, existing approaches remain slow, memory-intensive, and overly complex due to iterative optimization and dense per-Gaussian feature assignments. To address this, we propose LightSplat, a fast and memory-efficient training-free framework that injects compact 2-byte semantic indices into 3D representations from multi-view images. By assigning semantic indices only to salient regions and managing them with a lightweight index-feature mapping, LightSplat eliminates costly feature optimization and storage overhead. We further ensure semantic consistency and efficient inference via single-step clustering that links geometrically and semantically related masks in 3D. We evaluate our method on LERF-OVS, ScanNet, and DL3DV-OVS across complex indoor-outdoor scenes. As a result, LightSplat achieves state-of-the-art performance with up to 50-400x speedup and 64x lower memory, enabling scalable language-driven 3D understanding. For more details, visit our project page https://vision3d-lab.github.io/lightsplat/.