Extract music features and label your favorite songs for further model training.
# open a nix shell with required dependencies
nix develop
# Launch a minimal test case with default parameters
./src/test.py
src/ffmpeg_spectrogram.sh is an unused relic.
- Clean up TODOs
- Refactor MFExtractor class for tight CLI integration. (Future used as a helper script by some super structure.)
- Determine location of storage. XDG? Package-local? Global state?
- Reset training data. By age/genre/name/source/ Plus renew/overwrite flag.
- Fully 'MFExtractor' class init by CLI passed args. Seed the members of @dataclass
- All this AI stuff man.
- Select only the most promising music features to train the model.
- Scramble input music or random noise as negative test against the models?
- VLC integration?