Danceable
Project completed May 10, 2018
Danceable is an app for finding music suitable for your favorite dance style. Users can select a song from Spotify, Youtube or SoundCloud, or upload an MP3. The song is then converted into an image called a spectrogram, which is then sliced and passed through a Convolutional Neural Network (CNN). The CNN classifies the song as suitable for any of the following dance styles:
- Afrobeat
- Bachata
- Brazilian Zouk
- Kizomba
Danceable also identifies the Title, Artist, Album, Year and BPM of the track. Users can set the correct BPM by tapping and submit corrections to the other meta-data.
Danceable builds off of Julien Despois’s work on using CNNs to classify music by genre (see Article and Repo).