Skip to content

Latest commit

 

History

History
40 lines (34 loc) · 1.8 KB

README.md

File metadata and controls

40 lines (34 loc) · 1.8 KB
Raw

A Practical Introduction to Automatic Audio Segmentation Using Deep Learning

In Proceedings of: Summer School: Deep Learning for Language Analysis - September, 2019

Goal

Train a deep neural network on VGGish audio features to automatically segment an audio file into speech and non-speech parts.

Setup

Please follow the instructions below in order.

  1. Install and setup the following:
    1. docker
    2. git
  2. Clone this repository.
    • Or, download it if you don't have git.
  3. Start Docker on your machine.
    • Check in Docker settings if appropriate sharing of drive is setup
    • Check in Docker settings if appropriate CPU/RAM limits are setup
  4. Build a docker image (Terminal on Linux/macOS, or Powershell on Windows): 
    # change directory to the clone, if not already
# replace ~ below with where you cloned this repository
cd ~/UoC-ml-school-2019

# Build a docker image
docker build --tag uoc:2019 $PWD
  5. Run a docker container with the built image (serving a Jupyter Lab instance) 
    # assuming you are in UoC-ml-school-2019 directory

# run the created image, exposing Jupyter Lab port to 8888, 
# and mounting the current directory inside the container
docker run -it -p 8888:8888 -v $PWD/:/ml-school uoc:2019 jupyter lab
  6. Open the served Jupyter Lab instance in the browser by following instructions in the terminal.
    • There will be a URL in the terminal that you can copy and then paste in the browser.
  7. Check setup by opening and following the instructions 00-check-setup.ipynb.