通过动态特征归一化稳定流式视频几何
Stabilizing Streaming Video Geometry via Dynamic Feature Normalization
DyFN 通过动态调制潜变量特征的统计量（均值和方差）来稳定单目几何基础模型在流式输入下的尺度漂移，仅微调 2% 参数即可复用现有强模型，为视频深度估计和在线三维重建提供了一种轻量、因果的时序一致性方案。
原文依据 we trace this instability to its root cause: fluctuations in latent feature statistics, whose mean and variance directly determine the predicted depth's scale and shift
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论从流式 RGB 输入中估计一致的三维几何对于自动驾驶、具身智能和大规模重建等现实应用至关重要。尽管现代单目几何基础模型在单张图像上达到了很高的精度，但在连续输入上表现出严重的时序不一致性，尤其以尺度漂移为主导。通过有针对性的实证分析，我们追溯这一不稳定性至其根本原因：潜变量特征统计量的波动，其均值和方差直接决定了预测深度的尺度和偏移。基于这一洞察，我们引入了动态特征归一化（DyFN），一个轻量级、因果的循环模块，能够动态且鲁棒地调制特征统计量，以在时间上保持稳定的几何。我们通过仅微调 DyFN（仅增加 2% 的参数）来使强大的预训练单目几何模型适应流式输入，同时保持骨干网络冻结，从而在不损害单图像精度的情况下实现时序一致性。在四个基准上的大量实验表明，DyFN 有效消除了诸如分层不连续和位置抖动等时序伪影，并达到了最优的时序稳定性，相比先前的流式方法提升了高达 14%，甚至超越了更重的非因果视频基线。项目页面：https://shawlyu.github.io/DyFN
英文原文摘要 1390 字符
Consistent 3D geometry estimation from streaming RGB input is crucial for real-world applications such as autonomous driving, embodied AI, and large-scale reconstruction. While modern monocular geometry foundation models achieve strong single-image accuracy, they exhibit severe temporal inconsistency on continuous input, notably dominated by scale-shift drifting. Through targeted empirical analysis, we trace this instability to its root cause: fluctuations in latent feature statistics, whose mean and variance directly determine the predicted depth's scale and shift. Building on this insight, we introduce Dynamic Feature Normalization (DyFN), a lightweight, causal recurrent module that dynamically and robustly modulates feature statistics to maintain stable geometry over time. We adapt powerful pretrained monocular geometry models for streaming by finetuning only DyFN, a mere 2% additional parameters, while keeping the backbone frozen, thereby achieving temporal consistency without compromising single-image accuracy. Extensive experiments across four benchmarks show that DyFN effectively eliminates temporal artifacts such as disjointed layering and positional jitter, and achieves state-of-the-art temporal stability, improving over prior streaming methods by up to 14% and even outperforming heavier non-causal video baselines. Project page: https://shawlyu.github.io/DyFN