search

lanl / OpenFWI

A collection of codes with OpenFWI project
BSD 3-Clause "New" or "Revised" License
61 stars 10 forks source link

Model migration problem #6

Closed lizongzeng closed 7 months ago

lizongzeng commented 7 months ago

Hello,author, Your work is great.Can I use the model trained on these data sets to process the salt body data?For example, SEG salt

Thank you very much for your attention

hanchenwang commented 7 months ago

Hello, thanks for your interest. Unfortunately, the pre trained model cannot be directly applied to salt models because we did not include any salt body in our dataset. I would suggest to use other dataset that contains salt body structures to retrain, at least fine-tune or transfer learning, with our networks.

lizongzeng commented 7 months ago

Thank you very much for your reply.Is there open source real data available to test the effectiveness of models trained in openfwi?

hanchenwang commented 7 months ago

Sorry, but I do not know any open source dataset that is designed for salt body imaging. You may rescale SEG salt for a preliminary test with our model. However, for fine tuning the model, I would suggest you build a few samples, for example 1000, and start from our pretrained models. For retraining, you would probably need a larger dataset designed for the task.

hanchenwang commented 7 months ago

You may find some open source data on SEG open data platform or the EDX platform of DOE: https://edx.netl.doe.gov/ Remember you will need to scale the data to our model capacity.

lizongzeng commented 7 months ago

Thank you very much for your reply. Wish you good health and success in your work. Thank you again.