Training data for reconstructing current density from magnetic field (50 ± 10 μm standoff, 64x64 resolution)

This dataset contains 157,696 simulated magnetic field/current density/standoff distance sets, created to teach neural networks to solve the magnetic inverse problem. The data is divided into 154 sets of files (B, J, and standoff distance), with 1024 different configurations per file. Configurations alternate between straight and curved wires every 8 images. This dataset contains data with a 64x64 resolution, and standoff distances sampled from a normal distribution with μ=50 μm, σ=10 μm. Trained models, model training scripts (for MAGIC-UNet architecture), related experimental data, and validation data used in publication can be found at https://doi.org/10.7910/DVN/SD6PVP