SecklowSounds music/speech classifier

Requirements

System:

• python2 / python3 (preferred)
• python2-pip / python3-pip (preferred)
• ffmpeg (as backend to librosa)

pip packages

• virtualenv (strongly advised)
• pip install -r requirements.txt

Get the data

run ./download_all.sh to download all, also the audio files, or ./download_csv to download only the annotations.

Get the chunks that satisfy selection

run python select_chunks.py

Compute melgrams

run python compute_melgrams.py

Train model

run python train.py MODEL_NAME with MODEL_NAME one of the following:

• cnn
• small_cnn
• crnn

TODO