PhysGaia
收藏资源简介：
PhysGaia是一个新型的物理感知数据集，专为动态新视角合成（DyNVS）设计，包含结构化物体和非结构化物理现象。与现有主要关注照片真实感重建的数据集不同，PhysGaia旨在积极支持物理感知动态场景建模。该数据集提供了复杂的动态场景，其中多个物体之间存在丰富的相互作用，它们之间能够真实地碰撞并交换力量。此外，它还包含多种物理材料，如液体、气体、粘弹性物质和纺织品，超越了现有数据集中普遍存在的刚体。PhysGaia中的所有场景都严格按照物理定律真实生成，利用精心选择的特定材料物理求解器。为了能够对物理建模进行定量评估，我们的数据集提供了包括3D粒子轨迹和物理参数（如粘度）在内的基本真实信息。为了促进研究的采用，我们还提供了使用最先进的DyNVS模型与我们的数据集的必要集成管道，并报告了它们的结果。通过解决物理感知建模数据集的严重缺乏，PhysGaia将显著推动动态视图合成、基于物理的场景理解和与物理模拟集成的深度学习模型的研究。
PhysGaia is a novel physics-aware dataset designed for Dynamic Novel View Synthesis (DyNVS), encompassing both structured objects and unstructured physical phenomena. In contrast to existing datasets that primarily focus on photorealistic reconstruction, PhysGaia is intentionally developed to actively support physics-aware dynamic scene modeling. This dataset features complex dynamic scenes with rich interactions among multiple objects, where realistic collisions and force exchanges occur between entities. Furthermore, it incorporates a wide spectrum of physical materials including liquids, gases, viscoelastic materials and textiles, going beyond the rigid bodies that dominate most existing datasets. All scenes in PhysGaia are rigorously generated in accordance with physical laws, leveraging carefully selected material-specific physics solvers. To enable quantitative evaluation of physics modeling, our dataset provides ground-truth information including 3D particle trajectories and physical parameters such as viscosity. To facilitate research adoption, we also provide the necessary integration pipelines for deploying state-of-the-art DyNVS models with our dataset, and report their benchmark results. By addressing the critical shortage of physics-aware modeling datasets, PhysGaia will significantly advance research in dynamic view synthesis, physics-based scene understanding, and deep learning models integrated with physical simulation.
