Implementation of Generative Adversarial Networks, for Audio.
Idea is to be able to generate speech using GANs. Project inspired from keras-dcgan.
- GAN trained for generating/discriminating half second audios using data from Librispeech corpus and MLP models.
- Refactor the above code
- Parametrize training (model arch params, hyperparams)
- Implement GA for passing and evaling various model archs