FF_pred_DL_UDRM

Reproducible material for Efficient Upside-Down Rayleigh-Marchenko Imaging through Machine-Learned Focusing Function Estimation - Wang N., Ravasi M., Alkhalifah T.

1produce_data

• raymckimaging.ipynb: Calculating input and output data for NNs.

2test_data_NN

• inputdata: arrange for training, validation and test data: d+ d-
• outputdata: arrange for training, validation and test data: ff+ ff-
• create_mask.ipynb: calculate time mask

3train_NN_pred

• main.py : The entry point of this project. It coordinates the overall workflow by calling functions from the other sub-programs.
• arrange_data.py: Since the data is too large, we use this file to arrange for the total data.
• model.py: The architecture of the U-Net.
• train.py: Training process, validation process and predction process.

4imaging_ff

• Image_simple.ipynb: carry out UD-RM imaging

5produce_sparse_data

• simple_sparse.py: produce input and output data based on the sparse geometry to train the NNs.

6imaging_sparse

• Image_sparse.ipynb: carry out UD-RM imaging for sparse data

7produce_volve_data

• volve.py: produce input and output data for Volve field data to train the NNs.

8imaging_volve

• Image_volve.ipynb: carry out UD-RM imaging for Volve field data

Environment

To ensure reproducibility of the results, we have provided a environment-latest.yml file. Ensure to have installed Anaconda or Miniconda on your computer.

After that simply run:

./install_env-latest.sh

It will take some time, if at the end you see the word Done! on your terminal you are ready to go!

Cite us

Wang N. et al. (2026) Efficient Upside-Down Rayleigh-Marchenko Imaging through Machine-Learned Focusing Function Estimation. submitted to Geophysics.