An app to help you discover music that suits your dance style.
This app builds off of a Convolutional Neural Network designed to detect musical genres:
nihal223's github repo: https://github.com/nihal223/DeepAudioClassification/blob/master/README.md which was forked from despoisj's repo. The original methodology is described here: article on Medium
I modified nihal223's code to increase the sample size, label data based on dance-style directories instead of ID3 genre tags, and accommodate characters from foreign languages. I then built off of the model using a range of NPM packages to create a user-friendly application.
Download:
git clone
Place the following file in the /CNN: https://www.dropbox.com/s/c9g4uv27jzurbxc/musicDNN.tflearn.data-00000-of-00001?dl=0
Install:
If you don't have them, first install the latest versions of NodeJS and Python.
pip install libmagic eyed3 sox tensorflow tflearn
npm install
Run:
npm run start
Use at localhost:8000
- Select a dance style
- View playlists for inspiration
- Upload an mp3 file and click "Analyze MP3"
- See the results:
- listen to your track
- view its spectrogram
- see its BPM and meta data
- see what the CNN model thinks you can dance to it
- TAP to the beat while listening to set the BPM
- Check the styles you think it suits
- Fill in or edit the meta data
- Submit your feedback and download your corrected MP3