This repository acts as a summary of the work produced for the 2022 Univ of Washington eScience Institute Winter Incubator Program.
Project lead: Zoe Krauss, graduate student, School of Oceanography
eScience Liaison: Scott Henderson
The project description can be found here.
The workflow we developed accomplishes the following:
- downloads continuous offshore seismic data
- preprocesses the seismic data using either filtering or DeepDenoiser, a pretrained deep neural network denoiser
- extracts P- and S-wave picks from the continuous data using pretrained EQTransformer, a deep neural network phase picker
- evaluates the performance of the pre-trained network by comparison to a manually produced catalog
Our workflow makes extensive use of the Seisbench package to apply these pretrained models.
An example of running this workflow on offshore data from Alaska can be found in EXAMPLE.ipynb.
Resources to deploy this workflow on the cloud, using Azure, can be found in this repository.
The next step in this project is to retrain the machine learning networks using offshore data. The repository where this work is beginning is here (but is very much still in the development phase).