I was training facebook/denoiser model with interspeech2020/master dataset(dataset is generated with the default configuration). But it seems that in the case of non-stationary data it works really bad, although the author's code works correctly. I suspect that if there is any non-stationary noise in interspeech2020 or not. Or am I doing something wrong in the configuration file?
I was training facebook/denoiser model with interspeech2020/master dataset(dataset is generated with the default configuration). But it seems that in the case of non-stationary data it works really bad, although the author's code works correctly. I suspect that if there is any non-stationary noise in interspeech2020 or not. Or am I doing something wrong in the configuration file?