MAID: A deep learning dataset specifically designed for magnetic anomaly data interpolation tasks

# Overview The magnetic anomaly interpolation dataset (MAID) is a deep learning dataset specifically designed for magnetic anomaly data interpolation tasks. The original magnetic anomaly data are derived from the EMAG2v3 database. We selected eight distinct geographical regions and two different height planes of magnetic anomaly data in EMAG2v3 for dataset construction. Using the kriging method, we processed the original magnetic anomaly data to generate datasets with eight distinct grid spacings. Subsequently, each grid-spacing dataset was partitioned using fixed-size 256x256 windows. This process ultimately yielded MAID, which comprised 23501 magnetic anomaly samples. The dataset was partitioned into training, validation, and test sets containing 15387, 7139, and 975 samples, respectively. All data are stored in .mat format. # Details of the MAID dataset The details of the MAID dataset are shown in the table below. Region Altitude Longitude Latitude Grid spacing (m) Type of dataset Number Region 1 Sealevel 60°W~18°W 20°N~40°N 200, 400, 600, 800, 1000 Training set 4433 Region 2 Sealevel 162°W~132°W 10°N~45°N 200, 400, 600, 800, 1000 Training set 3883 Region 3 Sealevel 120°E~144°E 50°S~40°S 250, 500, 750, 1000 Test set 446 Region 4 Sealevel 30°W~0° 40°S~0° 200, 400, 600, 800, 1000 Validation set 4550 Region 5 4000 m 120°W~72°W 40°N~60°N 250, 500, 750, 1000 Training set 4172 Region 6 4000 m 18°E~60°E 50°N~65°N 200, 400, 600, 800, 1000 Training set 2899 Region 7 4000 m 120°E~150°E 30°S~20°S 200, 400, 600, 800, 1000 Test set 529 Region 8 4000 m 102°W~60°W 60°N~70°N 200, 400, 600, 800, 1000 Validation set 2589