A preliminary review of the efficacy of several acoustic autodetection algorithms to identify North Atlantic right whale calls, and recommendations for next steps to further assess and optimize these algorithms
Bibliographic record
Abstract
Automated detection and classification of the vocalizations of North Atlantic right whale (NARW) and other marine mammals is a highly desirable goal for researchers and managers seeking to monitor areas for whale presence as the basis to implement mitigation measures. Such automated acoustic processing is particularly important for real-time monitoring approaches where there are large-scale acoustic data inputs. All of the Detection and Classification Systems (DCSs) used by Fisheries and Oceans Canada (DFO) are expected to perform similarly well, given the metric (e.g., hours with calls/day) used to present NARW occurrence time-series. Previously, this was demonstrated by comparing performances of a variety of detectors during studies in 2004, 2013, and 2017. Spectroplotter (a commercial programme) and Low-Frequency Detection and Classification System (LFDCS), which are the two systems that have been used to analyse acoustic data in Newfoundland and Labrador (NL) and Maritimes regions, perform well; although in one small study the LFDCS detected more actual NARW upcalls than Spectroplotter, but also generated more false positives. DCS performance is influenced by multiple factors, including the ambient noise levels relative to the characteristics of the NARW upcalls, the location of the hydrophone, the characteristics of the recorder instrumentation, software settings and thresholds, and other contextual features, such as the presence of other species. The next generation of DCSs will incorporate context into their logic (e.g., presence of other marine mammals or abiotic sound sources and signal-to-noise ratio [SNR]). Algorithm comparisons are less crucial in the historic NARW analyses as the metrics in which the present detection results are presented at a large enough scale (“has there been NARW detected at this recorder location today?”) that slight differences in algorithm performance would be subsumed in the amalgamation and summation process. At smaller spatial and temporal sampling scales, differences in algorithm performance become more apparent. Thorough testing of the different DCSs being used in Atlantic Canada would require a series of manually validated acoustic datasets from a representative set of locations, time frames, seasons, and recording hardware. Such a DCS comparison would be a useful activity but would require agreed upon performance metrices and thresholds for the DCS.
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 | 0.008 | 0.017 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.002 |
| Bibliometrics | 0.005 | 0.003 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.003 | 0.003 |
| Open science | 0.003 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.003 |
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".