Haoran Zhou

@tangchenjun Not necessarily, I generated each file using the same scheme and raw data with random sampling for 12 times.

It may be a good idea to monitor your training process and spot the bottleneck.

I used TIMIT and other data sets should also work fine.

http://academictorrents.com/details/34e2b78745138186976cbc27939b1b34d18bd5b3

@nagakei05 @pramodhrachuri Thanks man

Sorry but I can't tell if it is because of overfitting or underfitting and you can use some training samples to see what is the problem. From my experience, 144k...

Yeah, the test script just uses 3 chunks of frames. You can change that by modifying the AudioSampleReader.gen() into a python generator and iterate through all the frames in audio_test.py.

@dearwind153 All the sample point are all 0 in your file. Guess some error happened and I am not very sure. You probably should run that step by step to...

The model doesn’t have the module to detect the number of speakers in a mixture, and another module to decide the matching order to concatenate the chunks of frames is...

Thanks for the reference. There is a rep about that https://github.com/khaotik/DaNet-Tensorflow. However, there are some bugs and it cannot generate reasonable separation and I am currently working on that. It...