SEA-Flow3D：通过空间向量采样与多尺度细化实现简化、高效且准确的场景流
SEA-Flow3D: Simplified, Efficient, and Accurate Scene Flow via Spatial Vector Sampling and Multi-scale Refinement
SVS 模块在匹配点局部邻域内联合采样 3D 坐标与相关体，得到带显式空间向量的方向感知相关表示，可直接替换 RAFT-3D 类框架中纯 2D 特征相关，为 3D 运动场优化注入几何先验。
原文依据 jointly samples 3D coordinates and correlation volumes within the local neighborhood of matched points
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深度辅助的场景流估计已取得快速进展，主流稠密框架（如 RAFT-3D）仍主要依赖二维特征相关来优化三维运动场，这阻碍了其有效利用三维结构先验，进而限制了其在复杂场景中的鲁棒性。本文提出 SEA-Flow3D，一个简单、高效且准确的稠密场景流估计框架。其核心是空间向量采样（SVS）模块，该模块在匹配点的局部邻域内联合采样三维坐标与相关体，生成带有显式空间向量的方向感知相关表示，为后续优化提供强有力的几何引导。遵循简洁高效的原则，SEA-Flow3D 采用 RAFT 风格的多尺度循环细化架构，将基于 RNN 的优化器与上下文引导的上采样相结合，以更少的迭代次数获得更高的精度。在 KITTI 和 Sintel 上的大量实验表明，SEA-Flow3D 在保持显著效率与轻量设计的同时达到了最优性能。代码页面：https://github.com/HanLingsgjk/SEAFLOW3D。
英文原文摘要 1206 字符
Although depth-assisted scene flow estimation has advanced rapidly, mainstream dense frameworks (e.g., RAFT-3D) still rely primarily on 2D feature correlations to optimize 3D motion fields, which hinders their ability to exploit 3D structural priors effectively and consequently limits robustness in complex scenes. We present SEA-Flow3D, a simple, efficient, and accurate framework for dense scene flow estimation. At its core lies a Spatial Vector Sampling (SVS) module that jointly samples 3D coordinates and correlation volumes within the local neighborhood of matched points, producing a direction-aware correlation representation with explicit spatial vectors and providing strong geometric guidance for subsequent optimization. Following the simplicity-and-efficiency principle, SEA-Flow3D adopts a RAFT-style multi-scale recurrent refinement architecture, integrating an RNN-based optimizer with context-guided upsampling to achieve higher accuracy with fewer iterations. Extensive experiments on KITTI and Sintel demonstrate that SEA-Flow3D achieves state-of-the-art performance while maintaining remarkable efficiency and a lightweight design. Code page: https://github.com/HanLingsgjk/SEAFLOW3D.