SeismicSense: Earthquake Event Detection and Analysis

This repository contains the codebase for SeismicSense, a tool designed for detecting and analyzing earthquake events, focusing on P and S phase arrivals.

Dataset

The model utilizes the STEAD Dataset, a curated dataset for seismic event detection.

Workflow

Preprocessing and Training

1. Preprocessing and Training
   Start with the SeismicSense_preprocess_n_train.py script:
   • Set mode="prepare" to preprocess the data.
   • Once preprocessing is complete, set mode="train" to train the model.
   • We have our custom layer to make the model run on MCU.

2. Testing the Original Model
   Use SeismicSense_test.py to evaluate the trained model on the test set.

Model Quantization

1. Quantize the Entire Model
   Use SeismicSensequant.py to quantize the trained model.

   • Test the quantized model with SeismicSensequant_test.py.

2. Split Models for Quantization
   If you prefer to work with split models:

   • Use Split.py to split the model.
   • Quantize the split models using SeismicSensequant_split.py.
   • Test the quantized split models with TestingSplit_quant.py.

On-device Deployment

Use the LightEQ_extension deployment code for on-device deployment (see: LightEQ GitHub Repository). Minor modifications will be required.

Repository Overview

• SeismicSense_preprocess_n_train.py: Handles data preparation and model training.
• SeismicSense_test.py: Tests the original trained model.
• SeismicSensequant.py: Quantizes the entire model.
• SeismicSensequant_test.py: Tests the fully quantized model.
• Split.py: Splits the model into parts for modular handling.
• SeismicSensequant_split.py: Quantizes the split models.
• TestingSplit_quant.py: Tests the quantized split models.

Getting Started

1. Clone this repository:

   git clone <repo_url>
   cd <repo_name>

2. Install the required dependencies:

   pip install -r requirements.txt

3. Follow the workflow steps above for your use case.

Acknowledgment

The STEAD Dataset: STEAD GitHub Repository