search

JUiscoming / GAN_decoded_audio_enhancement

The unofficial PyTorch Implementation of "Audio codec enhancement with generative adversarial networks." (20' ICASSP)
5 stars 0 forks source link

readme

GAN_decoded_audio_enhancement

The unofficial PyTorch Implementation of "Audio codec enhancement with generative adversarial networks." (20' ICASSP)

The spectral holes disappeared. but harmonic noise appeared.

The spectral normalization is used due to training stability.

Loss plot

Loss_Epochs

Loss_Iterations

Examples

Original waveform (p232_005.wav; VCTK)

p232_005_original

Decoded waveform

p232_005_decoded

Enhanced waveform

p232_005_enhanced