/USTC-Pickers

A Unified Set of seismic phase pickers Transfer learned for China

Primary LanguagePython

USTC-Pickers: a Unified Set of seismic phase pickers Transfer learned for China

Update summary for v0.1:

  1. The architecture of the USTC-Pickers is redesigned to exactly match the original PhaseNet written by W. Zhu&G. Beroza (2018). The earlier version of USTC-Pickers will not be maintained any more (i.e., PhaseNetLight, see this PR for details)
  2. We adopt diverse data augmentation techniques during training: adding data gaps, superimposing waveforms with the pure Noise from STEAD, randomly dropping 1 or 2 of the 3 components, waveform clipping, random waveform shifting.
  3. The CN picker trained with the whole DiTing data set can be directly accessed by SeisBench now, through sbm.PhaseNet.from_pretrained('diting').
  4. Many thanks to the users of USTC-Pickers, whose feedbacks make the new pickers less sensitive to background noise and thus significantly reduces the number of false picks.

1. Install Anaconda and requirements

  • Clone this repository to your device

    git clone https://github.com/JUNZHU-SEIS/USTC-Pickers.git
cd USTC-Pickers

  • Install a Python environment

    conda create -n USTC-Pickers
conda activate USTC-Pickers
pip install seisbench==0.5.2

2. Transfer-learned pickers

Located in the directory: USTC-Pickers/model_list/v0.1/

3. Batch prediction

Detailed in this Notebook

4. Other tutorials

5. Citation

If you find this toolkit helpful, please cite papers below: