/meow-piano

Code for project "meow-piano" in RevolutionUC 2022

Primary LanguageJavaScript

meow-piano

Code for the project "meow-piano" at RevolutionUC 2022

Try it out at this link, but we recommend you build your own service, as our server runs too slow :(

Introduction

Everyone loves cats, and everyone loves music. We hope to build an automated model to convert piano clips into cat singings, to produce very cute and interesting songs!

Introduction video: youtube

Project link: DevPost

Deployment

Deploy the backend

First you need to have Python. Then, install the following dependencies:

  1. PyTorch. Check out https://pytorch.org/
  2. install piano_transcription_inference using pip or conda
  3. download the model from the release (~120M), and put it at the folder ~/piano_transcription_inference_data/ (you may need to create it). ~ refers to your home directory. On Linux, it is usually /home/<your-username>; on Windows, it is usually C:\Users\<your-username>.
  4. install flask, flask_cors, gevent using pip or conda
  5. install librosa using pip or conda. Also make sure you have ffmpeg installed on your system.

Finally, enter the flask-backend folder, open a command prompt, and run python main.py.

If there's no graphic card on your device or if CUDA is not installed, change line 25 of main.py into device='cpu'. This could be two times slower than using CUDA.

Deploy the frontend

If you are deploying on a server, just put the frontend-d3 folder on your www root.

If you are deploying locally, then:

  1. Make sure node.js is installed. Check out https://nodejs.org/en/
  2. Make sure anywhere is installed. To install, open a command prompt, type npm install -g anywhere
  3. Enter the frontend-d3 folder, open a command prompt, and run anywhere

Also, please change the first line of js/ui.js to the correct backend address.