发表机构
Southeast University; Huazhong University of Science and Technology（东南大学; 华中科技大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
UniDynamics利用事件流与RGB融合，通过扩散框架生成未来4D动态场景，无需长历史，显式建模运动场，在VKitti2和DSEC上达到最先进性能。
AI 中文摘要
我们提出了UniDynamics，一种基于扩散的框架，用于从单个事件-RGB对生成未来的4D动态场景（RGB、深度和光流），无需像现有方法那样需要长历史或控制先验，同时显式建模未来运动场。核心思想是利用事件流为单RGB外推提供替代运动先验，并通过多模态建模在整个生成过程中强制几何和运动约束。具体而言，我们设计了一个事件潜在增强（ELE）模块，将事件潜在特征对齐并增强为可注入扩散的条件特征，提供稳健的初始运动先验和可靠的纹理/结构线索。我们进一步引入了一个嵌入多尺度U-Net中的感知动态空间（PDS），该空间解耦并自适应地交互深度和光流，同时持续向外观特征反馈约束，提高几何-运动一致性，以实现物理上合理且时空连贯的预测。在VKitti2和DSEC上的实验展示了最先进的性能，产生了高质量、时间连贯且4D一致的未来预测，尤其是在具有挑战性的高速运动模糊情况下。
英文摘要
We propose UniDynamics, a diffusion-based framework for future 4D dynamic scenes (RGB, depth, and optical flow) generation from a single event-RGB pair, without requiring long histories or control priors as in existing methods, while explicitly modeling future motion fields. The core idea is to leverage event streams to offer an alternative motion prior for single-RGB extrapolation, and to enforce geometric and motion constraints throughout generation via multimodal modeling. Specifically, we design an Event Latent Enhancement (ELE) module to align and enhance event latents into diffusion-injectable conditioning features, providing robust initial motion priors and reliable texture/structure cues. We further introduce a Perceptual Dynamics Space (PDS) embedded in the multi-scale U-Net, which decouples and adaptively interacts depth and flow while continuously feeding back constraints to appearance features, improving geometric-motion consistency for physically plausible and spatiotemporally coherent prediction. Experiments on VKitti2 and DSEC demonstrate state-of-the-art performance, producing high-quality, temporally coherent, and 4D-consistent future predictions, especially under challenging high-speed motion blur.
Comments19 pages, 6 figures, conference, code: https://github.com/KK-xi/Unidynamics
Journal refECCV2026
