Hydrophone Anomaly Detections at Ocean Network Canada’s Cabled Observatories
Bibliographic record
Abstract
As a partner of the Canadian Integrated Ocean Observing System (CIOOS) Building Bridges project, Ocean Networks Canada (ONC) is developing an open-source self-supervised machine learning algorithm to detect anomalies in passive acoustics data from hydrophones. This set of labelled data has been manually produced for both quality control and machine learning training purposes. The types of anomalies that have been labeled in this dataset are: Anomaly, Data Gap, Dropout, Engine Noise, Rain, Sensitivity, Tonal, and Unknown Feature. The intention is for the algorithm to detect anomalies as part of operational quality control processes at ONC. The passive acoustics data for this labelled dataset were collected by ONC hydrophones during the period between July 22, 2015 to November 13, 2024. Hydrophones are devices containing transducers that convert underwater sound waves into electrical signals. They are acoustic instruments that can process data while they are being collected to produce calibrated waveform data. Hydrophones are typically used to study vocalizations of marine mammals, ship traffic and ambient noise. The hydrophones were deployed by ONC on underwater fixed-position platforms in various locations, primarily in the Vancouver Island area of British Columbia. The sampled locations are: Burrard Inlet, Patricia Bay, Folger Deep, Barkley Canyon, Main Endeavour Field and the Strait of Georgia, as well as one location near Kitlineq/Victoria Island in Nunavut (Cambridge Bay). These anomaly detection labels for passive acoustic data may benefit those collecting and monitoring large volumes of hydrophone data. By using an anomaly detection algorithm, time consumption for evaluating hydrophone data is reduced immensely. Instead of scanning all spectral data, a data specialist only needs to review the spectrograms that were flagged as anomalous. This project is conducted as part of CIOOS’s Building Bridges project. Building Bridges is a project approach to accelerating the adoption of artificial intelligence (AI) in the ocean sector, with a focus on connecting not-for-profit organizations with the tools and information necessary to understand and implement the opportunities offered by AI. The project duration is from July 1, 2023 to March 31, 2026. ONC, one of the main partners in this project, is based at the University of Victoria in British Columbia. Through a collaboration of four national academic and not-for-profit partners across Canada, Building Bridges will address multiple components of the high-level artificial intelligence pipeline, which begins with having the knowledge and skills to identify and develop solutions for scientific questions or problems which may be solved with artificial intelligence. The ONC lead for the project is Drew Snauffer, with Vanessa Stewart and Piya Rashid as the Project Managers. Spencer Bialek is the Machine Learning/AI specialist who developed and documented the hydrophone anomaly machine learning algorithm. He is working with Brendan Smith, the Passive Acoustics and AI Specialist. Herminio Foloni Neto, Alex Slonimer, and Jeannette Bedard are the Scientific Data Specialists, Lafranco Muzi is the Staff Scientist.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.002 | 0.003 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".