https://github.com/ncsoft/phonmatchnet
Official implementation of "PhonMatchNet: Phoneme-Guided Zero-Shot Keyword Spotting for User-Defined Keywords" (INTERSPEECH 2023)
https://github.com/ncsoft/phonmatchnet
Last synced: about 1 year ago
JSON representation
Official implementation of "PhonMatchNet: Phoneme-Guided Zero-Shot Keyword Spotting for User-Defined Keywords" (INTERSPEECH 2023)
- Host: GitHub
- URL: https://github.com/ncsoft/phonmatchnet
- Owner: ncsoft
- License: bsd-3-clause
- Created: 2023-05-23T04:54:41.000Z (about 3 years ago)
- Default Branch: main
- Last Pushed: 2024-06-03T00:57:33.000Z (about 2 years ago)
- Last Synced: 2025-04-08T20:51:36.824Z (about 1 year ago)
- Language: Python
- Homepage:
- Size: 4.3 MB
- Stars: 46
- Watchers: 3
- Forks: 8
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# :mag: PhonMatchNet: Phoneme-Guided Zero-Shot Keyword Spotting for User-Defined Keywords
Official implementation of PhonMatchNet: Phoneme-Guided Zero-Shot Keyword Spotting for User-Defined Keywords.
PyTorch version: https://github.com/ncsoft/PhonMatchNet/tree/pytorch
## Requirements
### Datasets
* [LibriPhrase](https://github.com/gusrud1103/libriphrase)
* [Google Speech Commands](https://arxiv.org/abs/1804.03209)
* [Qualcomm Keyword Speech](https://developer.qualcomm.com/project/keyword-speech-dataset)
Download the dataset and prepare it according to each guide.
## Getting started
### Environment
```bash
cd ./docker
docker build --tag udkws .
```
### Training
```bash
docker run -it --rm --gpus '"device=0,1"' \
-v /path/to/this/repo:/home/ \
-v /path/to/prepared/dataset:/home/DB \
ukws \
/bin/bash -c \
"python train.py \
--epoch 100 \
--lr 1e-3 \
--loss_weight 1.0 1.0 \
--audio_input both \
--text_input g2p_embed \
--stack_extractor \
--comment 'user comments for each experiment'"
```
### Monitoring
```bash
tensorboard --logdir ./log/ --bind_all
```
## CONTRIBUTING
Please post bug reports and new feature suggestions to the Issues and Pull requests tabs of this repo.