Skip to content

Commit

Permalink
Merge pull request #54 from jaehyun-ko/main
Browse files Browse the repository at this point in the history
update readme for download DNS challenge 2020 dataset
  • Loading branch information
jzi040941 authored Jan 22, 2023
2 parents 1b83df7 + e3fcdb8 commit 8ffae43
Show file tree
Hide file tree
Showing 5 changed files with 27 additions and 9 deletions.
1 change: 1 addition & 0 deletions .gitattributes
Original file line number Diff line number Diff line change
@@ -0,0 +1 @@
*.wav filter=lfs diff=lfs merge=lfs -text
13 changes: 12 additions & 1 deletion .gitignore
Original file line number Diff line number Diff line change
Expand Up @@ -145,4 +145,15 @@ src/*.pcm
*.wav

#binary folder
bin/
bin/

#dataset folder
sampledata/
test_input.pcm
test_output.pcm
test.output
training_set_sept12_500h
sampledata_vctk_DEMAND
training.h5

DNS-Challenge
1 change: 1 addition & 0 deletions DNS-Challenge
Submodule DNS-Challenge added at 70f192
21 changes: 13 additions & 8 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,14 @@ https://www.researchgate.net/publication/343568932_A_Perceptually-Motivated_Appr
- Sox
- Python>=3.6
- Pytorch


## Prepare sampledata
1. download and sythesize data DNS-Challenge 2020 Dataset before excute utils/run.sh for training.
```shell
git clone -b interspeech2020/master https://github.com/microsoft/DNS-Challenge.git
```
2. Follow the Usage instruction in DNS Challenge repo(https://github.com/microsoft/DNS-Challenge) at interspeech2020/master branch. please modify save directories at DNS-Challenge/noisyspeech_synthesizer.cfg sampledata/speech and sampledata/noise each.

## Build & Training
This repository is tested on Ubuntu 20.04(WSL2)

Expand All @@ -38,7 +45,7 @@ cd ..

3. feature generation for training with sampleData
```
bin/src/percepNet sampledata/speech/speech.pcm sampledata/noise/noise20db.raw 4000 test.output
bin/src/percepNet sampledata/speech/speech.pcm sampledata/noise/noise.pcm 4000 test.output
```

4. Convert output binary to h5
Expand All @@ -47,8 +54,10 @@ python3 utils/bin2h5.py test.output training.h5
```

5. Training
```
python3 rnn_train.py
run utils/run.sh
```shell
cd utils
./run.sh
```

6. Dump weight from pytorch to c++ header
Expand All @@ -65,11 +74,7 @@ cd ..
bin/src/percepNet_run test_input.pcm percepnet_output.pcm
```

## SampleData

clean speech - VCTK 48k wav https://datashare.is.ed.ac.uk/handle/10283/2791 (clean_train_set)

noise data - DEMAND 48k wav https://zenodo.org/record/1227121#__sid=js0 (*.48k.zip)

## Acknowledgements
[@jasdasdf]( https://github.com/jasdasdf ), [@sTarAnna]( https://github.com/sTarAnna ), [@cookcodes]( https://github.com/cookcodes ), [@xyx361100238]( https://github.com/xyx361100238 ), [@zhangyutf]( https://github.com/zhangyutf ), [@TeaPoly](https://github.com/TeaPoly ), [@rameshkunasi]( https://github.com/rameshkunasi ), [@OscarLiau]( https://github.com/OscarLiau ), [@YangangCao]( https://github.com/YangangCao ), [Jaeyoung Yang]( https://www.linkedin.com/in/jaeyoung-yang-354b21146 )
Expand Down
Empty file modified sampledata/noise/noise.pcm
100755 → 100644
Empty file.

0 comments on commit 8ffae43

Please sign in to comment.