Cross-Domain Volume Database
收藏资源简介:
该数据集是一个大规模跨领域科学模拟体积数据库,由圣母大学研究团队构建,旨在为学习型体积压缩模型提供多样化训练数据。数据库包含6376个经过筛选的体积数据,源自21个不同科学领域的模拟项目,覆盖了天体物理、湍流燃烧等多种物理量和空间结构。数据集通过感知哈希技术进行去冗余处理,确保样本多样性以提升模型泛化能力。该数据库主要用于训练EVOLVE压缩框架,解决科学模拟数据在离线存储和传输中的高压缩比需求,支持跨领域泛化压缩任务。
This dataset is a large-scale cross-domain scientific simulation volumetric database, constructed by the research team at the University of Notre Dame, aiming to provide diverse training data for learned volumetric compression models. The database contains 6,376 curated volumetric datasets derived from simulation projects across 21 distinct scientific disciplines, covering various physical quantities and spatial structures such as astrophysics and turbulent combustion. The database utilizes perceptual hashing technology for redundancy removal, ensuring sample diversity to enhance the model's generalization capability. It is primarily used to train the EVOLVE compression framework, addressing the high compression ratio requirements of scientific simulation data during offline storage and transmission, and supporting cross-domain generalized compression tasks.




