Skip to content

v-ko/bg-open-speech-recording

 
 

Repository files navigation

Описание на fork-a

Този форк е адаптация на сайта за събиране на аудио данни за целите на машинното обучение, създаден от Pete Warden. Основните промени са свързани с енкодинга и преводи на текстовете.

Оригинално README:

Open Speech Recording is a small web application to collect short snippets of speech, and upload them to cloud storage. It's designed to help gather open speech data sets to train machine learning systems.

It's based around a small Flask app that will run on Google App Engine. This serves up a client-side Javascript app that prompts for a series of words, records the audio, and then POSTs the results back to the server.

Running

To get started, you'll need to edit app.yaml to point to your own storage bucket and update the session key. If you have the Google Cloud SDK set up, you should be able to run a local copy with this command:

dev_appserver.py app.yaml

I've often had trouble getting local copies of the app to work with cloud storage, so you may see errors on the final upload stage with this setup. To deploy it to an appspot instance, run this:

gcloud app deploy

Credits

Thanks to the Mozilla team for the Web Dictaphone sample application that I used as a starting point, Sole for the oscilloscope, and the Flask team for a lovely Python microframework!

Written by Pete Warden, [email protected].

About

Web application to record speech for an open data set

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • HTML 42.7%
  • JavaScript 37.8%
  • CSS 10.7%
  • Python 8.6%
  • Shell 0.2%