- pip install requirements.txt
- Download medium_orca.mdl from here.
- Put medium_orca.mdl in models/
- Run:
python3 process_wavs.py --path /path/to/data/
on folder of 2 second long snippets of whale/background sounds. This will create a folder /path/to/data/proccessed/ with resampled wavs and spectrograms in .npy format ready to be input into the model. - Run Train Model and Test SKRW cells in OrcaML.ipynb, making sure to update paths to point to the processed/ folder from step 3.
- Process data as in step 3. of Instructions for Testing.
- Place data into data/ structured as shown in the first cell of OrcaML.ipynb
- Uncomment training in Train Model cells of OrcaML.ipynb and run.