Initial commit
Browse files- README.md +73 -0
- pytorch_model.bin +3 -0
README.md
ADDED
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
tags:
|
3 |
+
- asteroid
|
4 |
+
- audio
|
5 |
+
- DCUNet
|
6 |
+
- audio-source-separation
|
7 |
+
datasets:
|
8 |
+
- Libri1Mix
|
9 |
+
- enh_single
|
10 |
+
license: cc-by-sa-3.0
|
11 |
+
inference: false
|
12 |
+
---
|
13 |
+
|
14 |
+
## Asteroid model `JorisCos/DCUNet_Libri1Mix_enhsignle_16k`
|
15 |
+
|
16 |
+
Description:
|
17 |
+
|
18 |
+
This model was trained by Joris Cosentino using the librimix recipe in [Asteroid](https://github.com/asteroid-team/asteroid).
|
19 |
+
It was trained on the `enh_single` task of the Libri1Mix dataset.
|
20 |
+
|
21 |
+
Training config:
|
22 |
+
|
23 |
+
```yml
|
24 |
+
data:
|
25 |
+
n_src: 1
|
26 |
+
sample_rate: 16000
|
27 |
+
segment: 3
|
28 |
+
task: enh_single
|
29 |
+
train_dir: data/wav16k/min/train-360
|
30 |
+
valid_dir: data/wav16k/min/dev
|
31 |
+
filterbank:
|
32 |
+
stft_n_filters: 1024
|
33 |
+
stft_kernel_size: 1024
|
34 |
+
stft_stride: 256
|
35 |
+
masknet:
|
36 |
+
architecture: Large-DCUNet-20
|
37 |
+
fix_length_mode: pad
|
38 |
+
n_src: 1
|
39 |
+
optim:
|
40 |
+
lr: 0.001
|
41 |
+
optimizer: adam
|
42 |
+
weight_decay: 1.0e-05
|
43 |
+
training:
|
44 |
+
batch_size: 2
|
45 |
+
early_stop: true
|
46 |
+
epochs: 200
|
47 |
+
gradient_clipping: 5
|
48 |
+
half_lr: true
|
49 |
+
num_workers: 4
|
50 |
+
```
|
51 |
+
|
52 |
+
|
53 |
+
Results:
|
54 |
+
|
55 |
+
On Libri1Mix min test set :
|
56 |
+
```yml
|
57 |
+
si_sdr: 13.154035391645971
|
58 |
+
si_sdr_imp: 9.704254085786271
|
59 |
+
sdr: 13.568058873121435
|
60 |
+
sdr_imp: 10.065396073908367
|
61 |
+
sar: 13.568058873121435
|
62 |
+
sar_imp: 10.065396073908367
|
63 |
+
stoi: 0.9199373340235417
|
64 |
+
stoi_imp: 0.12401751048300132
|
65 |
+
```
|
66 |
+
|
67 |
+
|
68 |
+
License notice:
|
69 |
+
|
70 |
+
This work "DCUNet_Libri1Mix_enhsignle_16k" is a derivative of [LibriSpeech ASR corpus](http://www.openslr.org/12) by Vassil Panayotov,
|
71 |
+
used under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/); of The WSJ0 Hipster Ambient Mixtures
|
72 |
+
dataset by [Whisper.ai](http://wham.whisper.ai/), used under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/) (Research only).
|
73 |
+
"DCUNet_Libri1Mix_enhsignle_16k" is licensed under [Attribution-ShareAlike 3.0 Unported](https://creativecommons.org/licenses/by-sa/3.0/) by Joris Cosentino
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:cf0e2d801a97f32fe2ea30d5808815714a0578a3b15f8d54d047033da0799ea2
|
3 |
+
size 39174203
|