RockNet

Rockfall and earthquake detection and association via multitask learning and transfer learning.<br /> Our preprint article can be found here.

Complete dataset

Please also download the complete data hosted on Dryad (https://doi.org/10.5061/dryad.tx95x6b2f), follow the instructions and place the files to specified directories in this repository.

Summary

• Installation
• Make prediction on hourly SAC files

Installation

To run this repository, we suggest Anaconda and pip for environment managements.

Clone this repository:

git clone https://github.com/tso1257771/RockNet.git
cd RockNet

Create a new environment

conda create -n rocknet python==3.7.3 anaconda
conda activate rocknet
pip install --upgrade pip
pip install -r ./requirements.txt --ignore-installed

Make prediction on hourly SAC files

In this repository, we provide two hourly three-component seismograms as examples for making predictions on continuous data.<br /> The data seismograms were collected in the Luhu tribe, Miaoli county, Taiwan.<br />

Enter the directory ./Luhu_pred_ex<br />

cd ./Luhu_pred_ex

1. Run script Luhu_pred_ex/P01_net_STMF.py to generate the output functions (also in SAC format) in Luhu_pred_ex/net_pred from the provided SAC files Luhu_pred_ex/sac<br />

python P01_net_STMF.py

2. Plot some prediction results<br />

python P02_plot.py