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
  conda install python
  pip install seisbench==0.3.0

2. Transfer-learned pickers

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

3. Batch prediction

Detailed in this Notebook

4. Other tutorials

• Video tutorials for a brief introduction and technical details
• Slide: 2022_ustc_seis_workshop.pptx

5. Citation

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

---

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

  Suggestions on how to use USTC-Pickers.

---

---

• SeisBench - A Toolbox for Machine Learning in Seismology

  Reference publication for software.

---

• Which picker fits my data? A quantitative evaluation of deep learning based seismic pickers

  Example of in-depth bencharking study of deep learning-based picking routines using the SeisBench framework.

---