SeisAug: A Comprehensive Data Augmentation Python Toolkit for Deep Learning

The SeisAug toolkit addresses a significant challenge in seismological studies: the limited availability of region- and depth-specific labeled data. This scarcity poses a preliminary drawback when applying deep learning and machine learning techniques. To enhance model performance, we propose data augmentation as a simple yet effective solution.

Getting Started

Clone the Repository Locally

To clone the repository to your local machine, use the following command:

git clone https://github.com/ISR-AIML/SeisAug.git

Running in Google Colab

1. Open a new notebook in Google Colab.
2. Execute each step individually in separate cells using the following commands:

!git clone https://github.com/ISR-AIML/SeisAug
%cd SeisAug
!pip install -r requirements.txt  # After restarting the session
%cd SeisAug
%run SeisAug.ipynb

Installation Via pip

pip install -r requirements.txt

Installation Via Conda

Choose one of the following options:

1. Create a Conda environment from the provided environment.yml file:

conda env create -f environment.yml

2. Alternatively, install the dependencies directly from the requirements.txt file:

conda install --file requirements.txt

Example mseed files provided in the Repo are from SCEDC

SCEDC (2013): Southern California Earthquake Data Center. Caltech. Dataset. doi:10.7909/C3WD3xH1.

Example of adding random noise at levels 20%, 40%, 60%, and 80% to the signal.

Example of adding non-periodic noise in the range of 15-30 Hz. Changes in frequency content of the added noise can be observed in the corresponding signal.

Example of adding periodic noise at 30 Hz, change in frequency content of added noise can be observed in the corresponding spectrograms.