Cloudy - a 3D Clouds Dataset
收藏资源简介:
Cloudy - a 3D Clouds Dataset是由慕尼黑工业大学的研究团队创建的一个包含1000个合成3D云密度场的数据集。该数据集通过JangaFX流体模拟器生成,模拟了云的形成和动态过程,捕捉了浮力、湍流和扩散等物理现象。数据集的分辨率为512×256×512,旨在为单视图3D重建任务提供高质量的云密度场数据。数据集的创建过程包括应用Perlin噪声函数和不同的粒子发射形状,以增加云的自然随机性和多样性。该数据集主要用于训练扩散模型,以解决单视图3D重建中的复杂光传输问题,特别是在云、烟雾等体积材料的重建中。通过该数据集,研究人员能够生成与观测数据一致且具有空间一致性的3D密度分布。
Cloudy – a 3D Clouds Dataset, developed by a research team at the Technical University of Munich, is a curated collection of 1000 synthetic 3D cloud density fields. Generated via the JangaFX fluid simulator, this dataset models cloud formation and dynamic processes, capturing core physical phenomena such as buoyancy, turbulence, and diffusion. Boasting a resolution of 512×256×512, it is designed to supply high-quality cloud density field data for single-view 3D reconstruction tasks. The dataset’s creation pipeline applies Perlin noise functions and diverse particle emission shapes to enhance the natural randomness and structural diversity of the simulated clouds. Primarily utilized for training diffusion models to resolve complex light transport challenges in single-view 3D reconstruction, this dataset is particularly tailored for the reconstruction of volumetric materials including clouds and smoke. With this dataset, researchers can generate spatially consistent 3D density distributions that align with real-world observational data.

- 1Light Transport-aware Diffusion Posterior Sampling for Single-View Reconstruction of 3D Volumes慕尼黑工业大学 · 2025年



