Code for "Classification of seismic events in the mainland of China based on spectrograms and model interpretability"

Version 3.0

This repository contains the complete code implementation for "Classification of seismic events in the mainland of China based on spectrograms and model interpretability".

The upload includes two main components:

1. ResWaveQuake_Classification.zip

Implementation of seismic event classification using ResWaveQuake encoder architecture. The framework performs 3-class classification on seismic waveform data using spectrogram-based analysis with PyTorch Lightning.

Key features:

• ResWaveQuake encoder for feature extraction
• 3-class classification: Natural earthquake events (EQ), explosion events (EP), and collapse events (CL)
• Event-level prediction with majority voting across components
• Automatic class balancing and comprehensive evaluation
• Support for multiple seismic data formats (.mseed, .seed, .SAC, etc.)
• Resume training capability from checkpoints

2. Code for Interpretability Analysis.zip (Version 3.0)

Comprehensive interpretability analysis tools for understanding model decisions and feature importance in seismic event classification.

Components include:

• Cumulative heatmap analysis experiments
• Occlusion-based interpretability tests
• Single-event interpretation analysis

Version 3.0 Updates:

• Corrected manuscript title
• Updated color schemes in epicentral distance-based cumulative interpretability figures to improve accessibility for color vision deficiency

Both packages include detailed README files with installation instructions and usage guidelines.