System:
- python2 / python3 (preferred)
- python2-pip / python3-pip (preferred)
- ffmpeg (as backend to librosa)
pip packages
- virtualenv (strongly advised)
pip install -r requirements.txt
run ./download_all.sh
to download all, also the audio files, or ./download_csv
to download only the annotations.
run python select_chunks.py
run python compute_melgrams.py
run python train.py MODEL_NAME
with MODEL_NAME one of the following:
cnn
small_cnn
crnn