How kmshivani05/earthquake-p-wave-detection-system-using-u-net-and-gaussian-labeling works
This repository is an offline seismic P-wave detection research system. Its primary workflow reads three-channel HDF5 records, preprocesses waveforms and creates Gaussian arrival targets, trains configurable U-Net++ models, then predicts arrivals and evaluates timing accuracy. Separate capabilities cover dataset auditing, waveform visualization, model/noise experiments, and an optional Bayesian magnitude-estimation extension. The graph follows documented capabilities and indexed imports; unsampled modules are included where their role is clear, but their internal call paths are not inferred.
Main entry points: predict.py Inference Engine · dataset.py HDF5 Dataset · train.py Training Coordinator · evaluate_model.py Evaluation Entry
Components 27
Seismic Data
- Dataset Inspection
scripts/inspect_dataset.py - Dataset Audit
src/earthquake_detection/data/dataset_audit.py - HDF5 Dataset
src/earthquake_detection/data/dataset.py - Gaussian Targets
src/earthquake_detection/data/gaussian_labels.py
Signal Processing
- Experiment Config
src/earthquake_detection/config.py - Bandpass Filter
src/earthquake_detection/preprocessing/filtering.py - Waveform Normalize
src/earthquake_detection/preprocessing/normalization.py - Trace Windowing
src/earthquake_detection/preprocessing/windowing.py - Displacement Features
src/earthquake_detection/preprocessing/displacement.py
Modeling and Training
- Training Entry
scripts/train_model.py - Training Coordinator
src/earthquake_detection/training/train.py - Training Losses
src/earthquake_detection/training/losses.py - Trainer
src/earthquake_detection/training/trainer.py - 1D U-Net++
src/earthquake_detection/models/unetplusplus_1d.py - Dilated Blocks
src/earthquake_detection/models/dilated_blocks.py - Self-Attention
src/earthquake_detection/models/attention.py
Prediction and Evaluation
- Prediction Entry
scripts/predict.py - Inference Engine
src/earthquake_detection/inference/predict.py - Arrival Decoder
src/earthquake_detection/inference/decoder.py - Evaluation Entry
scripts/evaluate_model.py - Detection Metrics
src/earthquake_detection/evaluation/metrics.py - Scientific Plots
src/earthquake_detection/visualization/plots.py
Research Extensions
- Noise Experiment Entry
scripts/run_noise_experiment.py - Noise Robustness
src/earthquake_detection/experiments/noise_robustness.py - Bayesian Magnitude
src/earthquake_detection/magnitude/mcmc_magnitude.py
Other components
- Researcher
- HDF5 Records
Connections 35
- Researcher → Dataset Inspection: inspects data
README.md - Dataset Inspection → Experiment Config: loads settings
scripts/inspect_dataset.py - Dataset Inspection → Dataset Audit: runs audit
scripts/inspect_dataset.py - HDF5 Records → HDF5 Dataset: provides records
src/earthquake_detection/data/dataset.py - HDF5 Dataset → Bandpass Filter: filters traces
src/earthquake_detection/data/dataset.py - HDF5 Dataset → Waveform Normalize: normalizes traces
src/earthquake_detection/data/dataset.py - HDF5 Dataset → Displacement Features: derives features
src/earthquake_detection/data/dataset.py - HDF5 Dataset → Gaussian Targets: generates targets
src/earthquake_detection/data/dataset.py - Researcher → Training Entry: trains model
README.md - Training Entry → Experiment Config: loads settings
scripts/train_model.py - Training Entry → Training Coordinator: starts training
scripts/train_model.py - Training Coordinator → HDF5 Dataset: creates splits
src/earthquake_detection/training/train.py - Training Coordinator → 1D U-Net++: builds model
src/earthquake_detection/training/train.py - Training Coordinator → Training Losses: builds loss
src/earthquake_detection/training/train.py - Training Coordinator → Trainer: runs trainer
src/earthquake_detection/training/train.py - 1D U-Net++ → Dilated Blocks: uses blocks
src/earthquake_detection/models/unetplusplus_1d.py - 1D U-Net++ → Self-Attention: uses attention
src/earthquake_detection/models/unetplusplus_1d.py - Researcher → Prediction Entry: requests prediction
README.md - Prediction Entry → Experiment Config: loads settings
scripts/predict.py - Prediction Entry → Inference Engine: requests prediction
scripts/predict.py - Inference Engine → Bandpass Filter: filters input
src/earthquake_detection/inference/predict.py - Inference Engine → Waveform Normalize: normalizes input
src/earthquake_detection/inference/predict.py - Inference Engine → Displacement Features: derives features
src/earthquake_detection/inference/predict.py - Inference Engine → 1D U-Net++: builds model
src/earthquake_detection/inference/predict.py - Inference Engine → Arrival Decoder: decodes scores
src/earthquake_detection/inference/predict.py - Researcher → Evaluation Entry: evaluates model
README.md - Evaluation Entry → HDF5 Dataset: loads records
scripts/evaluate_model.py - Evaluation Entry → Inference Engine: predicts arrivals
scripts/evaluate_model.py - Evaluation Entry → Detection Metrics: computes metrics
scripts/evaluate_model.py - Evaluation Entry → Scientific Plots: plots results
scripts/evaluate_model.py - Researcher → Noise Experiment Entry: runs robustness study
README.md - Noise Experiment Entry → Noise Robustness: runs experiment
scripts/run_noise_experiment.py - Noise Robustness → Inference Engine: tests predictions
src/earthquake_detection/experiments/noise_robustness.py - Noise Robustness → Detection Metrics: scores results
src/earthquake_detection/experiments/noise_robustness.py - Researcher → Bayesian Magnitude: explores extension
README.md