metadata
tags:
- asteroid
- audio
- ConvTasNet
datasets:
- Libri3Mix
- sep_clean
license: CC BY-NC 4.0
inference: false
Asteroid model JorisCos/ConvTasNet_Libri3Mix_sepclean_16k
Description:
This model was trained by Joris Cosentino using the librimix recipe in Asteroid.
It was trained on the sep_clean task of the Libri3Mix dataset.
Training config:
data:
n_src: 3
sample_rate: 16000
segment: 3
task: sep_clean
train_dir: data/wav16k/min/train-360
valid_dir: data/wav16k/min/dev
filterbank:
kernel_size: 32
n_filters: 512
stride: 16
masknet:
bn_chan: 128
hid_chan: 512
mask_act: relu
n_blocks: 8
n_repeats: 3
n_src: 3
skip_chan: 128
optim:
lr: 0.001
optimizer: adam
weight_decay: 0.0
training:
batch_size: 8
early_stop: true
epochs: 200
half_lr: true
num_workers: 4
Results :
On Libri3Mix min test set :
'si_sdr': 8.932601610824145,
'si_sdr_imp': 12.299341066588594,
'sdr': 9.557260814240447,
'sdr_imp': 12.76957128385349,
'sir': 17.387646884037455,
'sir_imp': 20.599955591768484,
'sar': 10.686885056960504,
'sar_imp': -55.8894643263213,
'stoi': 0.8481258332025354,
'stoi_imp': 0.25528367853750356
License notice:
This work "ConvTasNet_Libri3Mix_sepclean_16k" is a derivative of LibriSpeech ASR corpus by Vassil Panayotov, used under CC BY 4.0. "ConvTasNet_Libri3Mix_sepclean_16k" is licensed under Attribution-ShareAlike 3.0 Unported by Cosentino Joris.