A deep learning music platform
- Create backend dataflow to pull mp3 and metadata from spotify api to a local dir
- Run data collection on enough playlists to gather ~10k songs
- Convert all mp3s into spectograms and save them locally
- Train a deep convolutional autoencoder on spectograms to generate ml encodings
- Store encodings and metadata in DB (SQL/SPARK/etc.)
- Run basic clustering on only spotify metadata
- Run clustering on only spectograms and combined to see best results
- TBD
Current backend environment setup
pip install spotipy
pip install keras
pip install tensorflow
python3 data_collection.py
TBD
This project is licensed under the MIT License - see the LICENSE.md file for details