This repository is for the paper "Synthesis of Realistic ECG using Generative Adversarial Networks".
The current files uploaded are for implementing Minibatch Discrimination (MBD) for a 2 Layer CNN discriminator, please note that for ECG data with MBD layers the training does not converge.
You can edit the Model.py file accordingly to remove MBD layers and/or to add more Convolution-Pooling layer as described in the paper.
Usage:
$python3 train.py
If you find this repo helpful in any way please cite our arXiv preprint:
@misc{delaney2019synthesis,
title={Synthesis of Realistic ECG using Generative Adversarial Networks},
author={Anne Marie Delaney and Eoin Brophy and Tomas E. Ward},
year={2019},
eprint={1909.09150},
archivePrefix={arXiv},
primaryClass={eess.SP}
}