One-dimensional (1D) Deep learning (DL) inversion of loop-loop electromagnetic induction (EMI) data using convolutional neural network.

This is the companion Python code of the paper by Moghadas GJI 2020 (see reference below).

This code contains the following scripts:

DLINVEMI_1D_Training: this code contains the main CNN algorithm for training EMI data. To train the network, 20,000 subsurface models were randomly generated considering 12 layers with conductivity range between 1-100 mS/m.

DLINVEMI_1D_Predictions: this code applies the trained CNN network on the EMI data (Transect 1 in the paper) measured from the Chicken Creek catchment (Brandenburg, Germany).

Reference

Moghadas, D., 2020, One-dimensional deep learning inversion of electromagnetic induction data using convolutional neural network, Geophysical journal international, DOI: 10.1093/gji/ggaa161

License

Apache

Contact

Davood Moghadas (moghadas@b-tu.de)