Datasets for SPNet

The datasets are used in paper Deep learning for locating contaminated sites from self-potential signals: a laboratory perspective.

SPtest.mat: This mat file includes 100,000 SP sources models and associated forward modeling SP signals. Note that the SP signals in this file have been added 1% to 3% Gaussian noise and have been normalized to [0,1]. Therefore, the datasets can be fed in SPNet training directly. If you want to have the actual forward modeling SP response, you can use the 3D forward modeling codes at Github https://github.com/geohang/SPNet.

Observed.mat: This is the laboratory-measured SP signal, which has been interpolated to have the same size as the input of the SPNet.

True.mat: The true locations of the contaminated source in the laboratory experiment.