Rapid Earthquake Phase analysis of Ocean-bottom, Regional and Teleseismic events with Deep Learning | REPORT DL
The SeisBench toolbox for machine learning in seismology was established by this project. This toolbox is both a collection of benchmark datasets, both from the literature and assembled by the project, and an open source Python module that provides a unified API for a variety of machine learning models as well as the benchmark datasets. This framework makes it both easy to apply state-of-the-art machine learning models to new datasets for seismologists without machine learning expertise, and also makes it easy for AI experts to benchmark new algorithms and machine learning models against existing ones for a whole range of datasets. Within this project, we carried out a first comprehensive comparison of machine learning detection and picking algorithms across many datasets. Further, specialised machine learning picker models were developed for ocean bottom data and depth phases.
The SeisBench framework has already found widespread adoption. The software repository had nine contributors, of which four were external (at the end of 2022). Furthermore, multiple teams have released datasets in the benchmark format defined by SeisBench.