Seismogram waveform datasets for ConvNetQuake_INGV

This data archive contains the training, validation and test datasets for ConvNetQuake_INGV as presented in this article:

Lomax, A., Michelini, A., & Jozinović, D. (2019). An Investigation of Rapid Earthquake Characterization Using Single‐Station Waveforms and a Convolutional Neural Network. Seismological Research Letters, 90(2A), 517–529. https://doi.org/10.1785/0220180311

The training [validation] datasets consist of 15,200 [1773] event and 10,724 [1198] noise three-component waveforms and associated metadata from MedNet stations using events from 2010 to 2018 at 0°–180° with lower magnitude limits set as a function of event epicentral distance.
The test datasets consists of 1003 event and 621 noise three-component waveforms and associated metadata and an extended set of 4074 test events from 2007 to 2009, selected otherwise with the same criteria as for the training and validation datasets.
 

The easiest way to work with the hdf5 file is to use the python library h5py. See README.txt