Evaluating the performance of the MIRFEE classifier plugin for PAMGuard at differentiating between whale vocalizations and anthropogenic noise in the Salish Sea
Bibliographic record
Abstract
The Music Information Retrieval Feature-Extracting Ensemble (MIRFEE) classifier was developed as a plugin for PAMGuard to provide species classification for the Whistle and Moan Detector (WMD) module. When used on audio recorded by hydrophones deployed in and around the Salish Sea for the purpose of killer whale detection, the WMD is routinely triggered by humpback whale vocalizations and vessel noise. The MIRFEE classifier was thus developed as a means of reducing these false positives. MIRFEE works by extracting features from both detection metadata and audio clips taken from when detections occur. These features are subsequently used as training data for an ensemble learning model. Pre-recorded hydrophone audio from 12 different deployment locations in the Strait of Juan de Fuca and Southern Gulf Islands across all seasons were run through the WMD. Manually-annotated detections produced by one of three classes—killer whale calls, humpback whale calls, or anthropogenic or environmental noise—were arranged into 18 subsets, and corresponding audio was subsequently run through the MIRFEE Feature Extractor. The resulting feature vectors were used to create two training sets that used different audio clip lengths, and the cross-validation results of the consequent training models are discussed.
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 | 0.002 | 0.004 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| 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".