The first step is to take the event catalog, take all the picks from it associated with the Z component, and save them in a simpler CSV format:
python -m netquake parse_catalog <input-nordic-path> <output-csv-path>The second step is to only take the traces containing events and corresponding to the Z component:
python -m netquake filter_traces <input-csv-catalog-path> <trace-folder-path> <output-folder-path>The third step is to adapt the traces so they can have an easier input for the neural network, this involves applying a bandpass filter and normalizing the signals. Frequencies are passed in Hertz:
python -m netquake adapt_traces <trace-folder-path> <output-folder-path> <band-min-freq> <band-max-freq>The fourth step is to slice into windows and save them on different folders if they contain an event or not:
python -m netquake gen_windows <catalog-path> <eventfull-trace-folder-path> <destination-folder> <window-size-seconds> <event-position>The fifth step is to pack all the datasets into a pickle:
python -m netquake gen_dataset <output-pickle-path> <trace-folder-paths>For viewing the signals:
python -m netquake display_traces <trace-folder-path> <trace-count>To train a network:
python -m netquake run_experiment <config-json-file> <network-json-file>To use a trained network:
python -m netquake apply_network <network-h5-file> <filtered-mseed-file>The first step is to create the network we want to use using Tensorflow and save it in JSON format.
import tensorflow as tf
model = tf.keras.Sequential()
model.add(tf.keras.layers.Dense(128, activation='relu', input_shape=(3001,))
model.add(tf.keras.layers.Dense(16, activation='relu')
model.add(tf.keras.layers.Dense(16, activation='relu')
model.add(tf.keras.layers.Dense(32, activation='relu')
model.add(tf.keras.layers.Dense(32, activation='sigmoid')
model.to_json('networks/mynet.json')We create then an experiment configuration.
{
"dataset_path": "dataset/windows/",
"epochs": 10,
"result_path": "results/result.csv",
"network_path": "networks/trained.h5"
}Finally we can run the entire pipeline.
python -m netquake parse_catalog dataset/catalog.nordic dataset/catalog.csv
python -m netquake filter_traces dataset/catalog.csv dataset/mseed/ dataset/eventful_mseed/
python -m netquake adapt_traces dataset/eventful_mseed/ dataset/filtered_mseed/ 0.5 10
python -m netquake gen_windows dataset/catalog.csv dataset/filtered_mseed/ dataset/windows/ 30 0.25
python -m netquake run_experiment experiment.json networks/mynet.json