Jacob-Stevens-Haas/tehom

Machine Learning on Underwater Acoustics

Tehom - Machine Learning on Underwater Acoustics

This package facilitates the creation of machine learning training data on underwater acoustics. While the raw data is available from Marine Cadastre and Ocean Networks Canada (via the onc package), tehom tracks downloads that it conducts and exposes more useful queries for data exploration and sampling. e.g.:

"What hydrophones outside Vancouver have data during April-June, 2017?"

"Where have I downloaded time/geo overlapping data?"

"Give me acoustic snippets as numpy arrays, labeled with whether a container ship was close to the hydrophone that recorded the snippet."

Requirements

• sqlite3 available as a command line program
• Create Ocean Networks Canada account and get your API token here

How to:

example.ipynb shows the basic data access and navivgation. Before working with any ONC data, you must once run

python -m tehom save-token <token>

Useful additional commands to start with:

• tehom.download_ships
• tehom.download_acoustics
• tehom.downloads.get_ais_downloads
• tehom.downloads.get_onc_downloads
• tehom.downloads.get_audio_availability
• tehom.downloads.show_available_data (plotting, use Jupyter terminal)
• tehom.sample (in progress!)

In addition to save-token, when run as a module/CLI, tehom can also download data.

About

"Tehom" is the Hebrew word for abyss, specifically the dark, chaotic oceans from which order and the world emerged.