Update README.md
Browse files
README.md
CHANGED
@@ -3,23 +3,63 @@ tags:
|
|
3 |
- audacity
|
4 |
inference: false
|
5 |
---
|
6 |
-
This is an Audacity wrapper for the model, forked from the repository JorisCos/ConvTasNet_Libri3Mix_sepnoisy_16k
|
7 |
This model was trained using the Asteroid library: https://github.com/asteroid-team/asteroid.
|
8 |
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
- audacity
|
4 |
inference: false
|
5 |
---
|
6 |
+
This is an Audacity wrapper for the model, forked from the repository `JorisCos/ConvTasNet_Libri3Mix_sepnoisy_16k`,
|
7 |
This model was trained using the Asteroid library: https://github.com/asteroid-team/asteroid.
|
8 |
|
9 |
+
The following info was copied directly from `JorisCos/ConvTasNet_Libri3Mix_sepnoisy_16k`:
|
10 |
+
|
11 |
+
Description:
|
12 |
+
This model was trained by Joris Cosentino using the librimix recipe in [Asteroid](https://github.com/asteroid-team/asteroid).
|
13 |
+
It was trained on the `sep_noisy` task of the Libri3Mix dataset.
|
14 |
+
Training config:
|
15 |
+
```yml
|
16 |
+
data:
|
17 |
+
n_src: 3
|
18 |
+
sample_rate: 16000
|
19 |
+
segment: 3
|
20 |
+
task: sep_noisy
|
21 |
+
train_dir: data/wav16k/min/train-360
|
22 |
+
valid_dir: data/wav16k/min/dev
|
23 |
+
filterbank:
|
24 |
+
kernel_size: 32
|
25 |
+
n_filters: 512
|
26 |
+
stride: 16
|
27 |
+
masknet:
|
28 |
+
bn_chan: 128
|
29 |
+
hid_chan: 512
|
30 |
+
mask_act: relu
|
31 |
+
n_blocks: 8
|
32 |
+
n_repeats: 3
|
33 |
+
n_src: 3
|
34 |
+
skip_chan: 128
|
35 |
+
optim:
|
36 |
+
lr: 0.001
|
37 |
+
optimizer: adam
|
38 |
+
weight_decay: 0.0
|
39 |
+
training:
|
40 |
+
batch_size: 8
|
41 |
+
early_stop: true
|
42 |
+
epochs: 200
|
43 |
+
half_lr: true
|
44 |
+
num_workers: 4
|
45 |
+
```
|
46 |
+
|
47 |
+
Results:
|
48 |
+
On Libri3Mix min test set :
|
49 |
+
```yml
|
50 |
+
si_sdr: 5.926151147554517
|
51 |
+
si_sdr_imp: 10.282912158535625
|
52 |
+
sdr: 6.700975236867358
|
53 |
+
sdr_imp: 10.882972447337504
|
54 |
+
sir: 15.364110064569388
|
55 |
+
sir_imp: 18.574476587171688
|
56 |
+
sar: 7.918866830474568
|
57 |
+
sar_imp: -0.9638973409971135
|
58 |
+
stoi: 0.7713777027310713
|
59 |
+
stoi_imp: 0.2078696167973911
|
60 |
+
```
|
61 |
+
License notice:
|
62 |
+
This work "ConvTasNet_Libri3Mix_sepnoisy_16k" is a derivative of [LibriSpeech ASR corpus](http://www.openslr.org/12) by Vassil Panayotov,
|
63 |
+
used under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/); of The WSJ0 Hipster Ambient Mixtures
|
64 |
+
dataset by [Whisper.ai](http://wham.whisper.ai/), used under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/).
|
65 |
+
"ConvTasNet_Libri3Mix_sepnoisy_16k" is licensed under [Attribution-ShareAlike 3.0 Unported](https://creativecommons.org/licenses/by-sa/3.0/) by Joris Cosentino
|