Giters

drethage / speech-denoising-wavenet

A neural network for end-to-end speech denoising

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Bad denoising results on the provided model

kfmn opened this issue · comments

commented

While the provided model in sessions/001 works well on NSDTSEA test files, the results on my own noisy files (recorded in real conditions) are much worse.

What did I do:

  1. Cloned the entire project
  2. Put my files into the 'test' subdir of the project dir
  3. Run the command
    THEANO_FLAGS=optimizer=fast_compile,device=gpu python main.py --mode inference --config sessions/001/config.json --noisy_input_path test
  4. Took results from sessions/001/samples/samples_1

Large part of speech was suppressed, although SNR is not very low.
Maybe I do something wrong?

To try it yourself: https://drive.google.com/open?id=1njlPLNjbTuY1QlW_19y06a1ywuImBUHo

commented

Same bad results with the pretrained model, even for the NSDTSEA test files. :(.

commented

Hi, kfmn,

I want to know the parameter size of wanenet? But i use the python 3, there are some problems on them. I just want to know the size of this model. Can you tell me the results of model.summary()?
Thank you very much!