tky823
tky823
`linear dense layer + overlap-add` is the same as a transposed convolution. https://github.com/kaituoxu/Conv-TasNet/blob/94eac1023eaaf11ca1bf3c8845374f7e4cd0ef4c/src/conv_tasnet.py#L140-L141
I also have the similar results; SDRi is 13.49, SI-SNRi is 12.94. And separated sources seem to be so distorted compared to audio samples by the author.
@adiyoss Thanks for your advice. I didn't notice comment out.
I'm not sure of # channels before frequency concatenation. The # of channels depends on the growth rate and # of D2 blocks. I added bottleneck convolution so that both...
What needs to be fixed - [x] multi dilated convolution - [x] timing of batch normalization - [x] # of output channels of D2 block - [x] order of D3...
Now, I updated D3Net architecture.
Hi, @lyghter. I'm now writing the training code. I am not sure if it will be available soon, but I plan to add it.
I'm not sure how my implementation of D3Net will work, so I don't know if I'll be able to participate anytime soon. If I can help, I will join your...
@lyghter I'm sorry I couldn't help you. Now, I'm sharing the scripts and results in [`egs/musdb18/d3net`](https://github.com/tky823/DNN-based_source_separation/tree/main/egs/musdb18/d3net).
Hi @mpariente, I'm interested in DNN-based source separation. I would be happy to contribute to Asteroid. What is the first step?