SuperBench
收藏资源简介:
SuperBench是一个专为科学机器学习中超分辨率技术设计的基准数据集,由劳伦斯伯克利国家实验室等机构创建。该数据集包含高达2048×2048分辨率的数据,涵盖流体流动、宇宙学和天气等多个科学领域。数据集的创建旨在通过提供标准化的评估和比较,推动超分辨率方法在科学任务中的应用。SuperBench不仅用于验证空间超分辨率性能,还评估了数据降级任务的鲁棒性。该数据集的应用领域包括加速数值模拟和提高天气、气候及相关领域的预测精度。
SuperBench is a benchmark dataset specifically designed for super-resolution techniques in scientific machine learning, developed by Lawrence Berkeley National Laboratory and other institutions. The dataset contains data with a maximum resolution of 2048×2048, covering multiple scientific fields including fluid dynamics, cosmology, and meteorology. It was created to advance the application of super-resolution methods in scientific tasks by providing standardized evaluation and comparative frameworks. SuperBench is used not only to validate spatial super-resolution performance but also to assess the robustness of data degradation tasks. Its application areas include accelerating numerical simulations and improving prediction accuracy in weather, climate, and related domains.

- 1SuperBench: A Super-Resolution Benchmark Dataset for Scientific Machine Learning劳伦斯伯克利国家实验室 · 2023年



