Data from "Cross-Domain Foundation Model Adaptation: Pioneering Computer Vision Models for Geophysical Data Analysis"

This dataset includes the collected geoscientific data of lunar images, seismic data, and DAS arrays.

It can be used to test the foundation model adaptation in the geophysical domain.

Below are some brief desription of the datasets:

1) Lunar images for crater detection[CAS] : 1000 are used for training and 199 for testing, each with a size of 1022x1022.

2) Seismic data for geobody identification[TGS] :  3000 are used for training and 1000 for testing, each with a size of 224x224.

3) Seismic data for facies classification[SEAM] :  250 are used for training and 45 for testing, each with a size of 1006x782.

4) Seismic data for deep fault detection :  1081 are used for training and 269 for testing, each with a size of 896x896.

5) DAS arrays for seismic event detection[Biondi] :  115 are used for training and 28 for testing, each with a size of 512x512.

Tips: All the images are “float32” and the labels are "int8".

The test code has been published on GitHub: Cross-Domain-Foundation-Model-Adaptation.