OpenSWI

OpenSWI是一个用于表面波频散曲线反演的综合性基准数据集，由表面波反演数据准备（SWIDP）流程生成。OpenSWI包含两个合成数据集，分别为OpenSWI-shallow和OpenSWI-deep，以及一个用于评估模型泛化能力的OpenSWI-real真实世界数据集。OpenSWI-shallow基于2D地质模型数据集OpenFWI，包含超过2200万个1D速度剖面及其对应的基本模式相速度和群速度频散曲线，涵盖了广泛的浅层地质结构。OpenSWI-deep由14个全球和区域3D地质模型组成，包含约126万个高保真的1D速度-频散数据对，适用于深层地球研究。OpenSWI-real从开源项目中编译，包含两组观测频散曲线及其对应的1D参考模型，用于评估深度学习模型的泛化能力。

OpenSWI is a comprehensive benchmark dataset for surface wave dispersion curve inversion, generated by the Surface Wave Inversion Data Preparation (SWIDP) workflow. It includes two synthetic datasets, namely OpenSWI-shallow and OpenSWI-deep, as well as an OpenSWI-real real-world dataset for evaluating model generalization capability. OpenSWI-shallow is based on the 2D geological model dataset OpenFWI, containing over 22 million 1D velocity profiles and their corresponding fundamental-mode phase and group velocity dispersion curves, covering a wide range of shallow geological structures. OpenSWI-deep consists of 14 global and regional 3D geological models, containing approximately 1.26 million high-fidelity 1D velocity-dispersion data pairs, suitable for deep Earth research. OpenSWI-real is compiled from open-source projects, including two sets of observed dispersion curves and their corresponding 1D reference models, which is used to evaluate the generalization ability of deep learning models.