Stabilizing Streaming Video Geometry via Dynamic Feature Normalization

• URL: http://arxiv.org/abs/2605.25308v1
• Date: Mon, 25 May 2026 00:13:15 GMT
• Title: Stabilizing Streaming Video Geometry via Dynamic Feature Normalization
• Abstract summary: Dynamic Feature Normalization (DyFN) is a causal recurrent module that modulates feature statistics to maintain stable geometry over time.<n>DyFN achieves state-of-the-art temporal stability, improving over prior streaming methods by up to 14% and even outperforming heavier non-causal video baselines.
• Score: 45.91399406203578
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: 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

Related papers

• Deformable Gaussian Occupancy: Decoupling Rigid and Nonrigid Motion with Factorized Distillation [81.44955493339835]
  DeGO is a deformable Gaussian occupancy framework that unifies decoupled Gaussian deformation with factorized 4D foundation-model distillation.<n> Experiments on the Occ3D-NuScenes benchmark demonstrate that our method achieves state-of-the-art performance under weak supervision.
  arXiv  Detail & Related papers  (2026-05-27T15:09:55Z)
• Physics-Informed Temporal U-Net for High-Fidelity Fluid Interpolation [0.0]
  We propose a novel Temporal U-Net architecture that integrates a VGG-based perceptual loss along with a Physics-Informed Bridge to overcome these issues.<n> Experimental results on multi-channel RGB fluid data show that our method clearly outperforms standard models.
  arXiv  Detail & Related papers  (2026-04-25T16:36:52Z)
• Robot-DIFT: Distilling Diffusion Features for Geometrically Consistent Visuomotor Control [31.523644271297616]
  A key bottleneck in generalizable robot manipulation is a structural mismatch between current visual backbones and the physical requirements of closed-loop control.<n>We propose Robot-DIFT, a framework that decouples the source of geometric information from the process of inference via Manifold Distillation.<n>By distilling a frozen diffusion teacher into a deterministic Spatial-Semantic Feature Pyramid Network (S2-FPN), we retain the rich geometric priors of the generative model while ensuring temporal stability, real-time execution, and robustness against drift.
  arXiv  Detail & Related papers  (2026-02-12T13:30:24Z)
• OLR-WAA: Adaptive and Drift-Resilient Online Regression with Dynamic Weighted Averaging [7.146027549101716]
  "OLR-WAA: An Adaptive and Drift-Resilient Online Regression with Dynamic weighted average"<n>This paper introduces "OLR-WAA: An Adaptive and Drift-Resilient Online Regression with Dynamic weighted average"
  arXiv  Detail & Related papers  (2025-12-14T17:39:51Z)
• LyTimeT: Towards Robust and Interpretable State-Variable Discovery [7.505092370141079]
  LyTimeT is a framework for interpretable variable extraction.<n>It learns robust and stable latent representations of dynamical video systems.<n>Our results demonstrate that combiningtemporal attention with stability constraints yields predictive models.
  arXiv  Detail & Related papers  (2025-10-22T16:03:10Z)
• Solving Inverse Problems with FLAIR [68.87167940623318]
  We present FLAIR, a training-free variational framework that leverages flow-based generative models as prior for inverse problems.<n>Results on standard imaging benchmarks demonstrate that FLAIR consistently outperforms existing diffusion- and flow-based methods in terms of reconstruction quality and sample diversity.
  arXiv  Detail & Related papers  (2025-06-03T09:29:47Z)
• STATIC : Surface Temporal Affine for TIme Consistency in Video Monocular Depth Estimation [14.635179908525389]
  Video monocular depth estimation is essential for applications such as autonomous driving, AR/VR, and robotics.<n>Recent transformer-based single-image monocular depth estimation models perform well on single images but struggle with depth consistency across video frames.<n>We propose STATIC, a novel model that learns temporal consistency in static and dynamic area without additional information.
  arXiv  Detail & Related papers  (2024-12-02T03:53:33Z)
• Towards Stabilized and Efficient Diffusion Transformers through Long-Skip-Connections with Spectral Constraints [51.83081671798784]
  Diffusion Transformers (DiT) have emerged as a powerful architecture for image and video generation, offering superior quality and scalability.<n>DiT's practical application suffers from inherent dynamic feature instability, leading to error amplification during cached inference.<n>We propose Skip-DiT, an image and video generative DiT variant enhanced with Long-Skip-Connections (LSCs) - the key efficiency component in U-Nets.
  arXiv  Detail & Related papers  (2024-11-26T17:28:10Z)
• EmerNeRF: Emergent Spatial-Temporal Scene Decomposition via Self-Supervision [85.17951804790515]
  EmerNeRF is a simple yet powerful approach for learning spatial-temporal representations of dynamic driving scenes. It simultaneously captures scene geometry, appearance, motion, and semantics via self-bootstrapping. Our method achieves state-of-the-art performance in sensor simulation.
  arXiv  Detail & Related papers  (2023-11-03T17:59:55Z)
• SceNeRFlow: Time-Consistent Reconstruction of General Dynamic Scenes [75.9110646062442]
  We propose SceNeRFlow to reconstruct a general, non-rigid scene in a time-consistent manner. Our method takes multi-view RGB videos and background images from static cameras with known camera parameters as input. We show experimentally that, unlike prior work that only handles small motion, our method enables the reconstruction of studio-scale motions.
  arXiv  Detail & Related papers  (2023-08-16T09:50:35Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.