msb-public
/
PaddleSpeech
mirror of https://github.com/PaddlePaddle/PaddleSpeech
1
0
Code Issues Projects Releases Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from '368e3e1b59'
${ noResults }
PaddleSpeech/examples/hey_snips/kws0
History
KP bf056c013d
Refactor paddleaudio to paddlespeech.audio 		3 years ago
..
conf Refactor paddleaudio to paddlespeech.audio 3 years ago
local Update KWS example. 3 years ago
README.md Add KWS example. 3 years ago
path.sh Add KWS example. 3 years ago
run.sh Update KWS example. 3 years ago

README.md

MDTC Keyword Spotting with HeySnips Dataset

Dataset

Before running scripts, you MUST follow this instruction to download the dataset: https://github.com/sonos/keyword-spotting-research-datasets

After you download and decompress the dataset archive, you should REPLACE the value of data_dir in conf/*.yaml to complete dataset config.

Get Started

In this section, we will train the MDTC model and evaluate on "Hey Snips" dataset.

CUDA_VISIBLE_DEVICES=0,1 ./run.sh conf/mdtc.yaml

This script contains training and scoring steps. You can just set the CUDA_VISIBLE_DEVICES environment var to run on single gpu or multi-gpus.

The vars stage and stop_stage in ./run.sh controls the running steps:

  • stage 1: Training from scratch.
  • stage 2: Evaluating model on test dataset and computing detection error tradeoff(DET) of all trigger thresholds.
  • stage 3: Plotting the DET cruve for visualizaiton.