Comparison of Three Individual Identification Algorithms for Sperm Whales ( <i>Physeter macrocephalus</i> ) after Automated Detection
Bibliographic record
Abstract
Abstract Photo-identification of individual sperm whales ( Physeter macrocephalus ) is the primary technique for mark-recapture-based population analyses for the species The visual appearance of the fluke — with its distinct nicks and notches — often serves as the primary visual differentiator, allowing humans to make recorded sightings of specific individuals. However, the advent of digital photography and the significant increase in volume of images from multiple projects in combination with pre-existing historical catalogs has made applying the method more challenging.with the required human labor for de-duplication (reduction of Type II errors) and reconciliation of sightings between large datasets too cost- and time-prohibitive. To address this, we trained and evaluated the accuracy of PIE v2 (a triplet loss network) along with two existing fluke trailing edge-matching algorithms, CurvRank v2 and Dynamic Time Warping (DTW), as a mean to speed comparison among a high volume of photographs. Analyzed data were collected from a curated catalog of well-known sperm whales sighted across years (2005-2018) off the island of Dominica. The newly-trained PIE model outperformed the older CurvRank and DTW algorithms, and PIE provided the following top-k individual ID matching accuracy on a standard min-3/max-10 sighting training data set: Rank-1: 87.0%, Rank-5: 90.5%, and Rank-12: 92.5%. An essential aspect of PIE is that it can learn new individuals without network retraining, which can be immediately applied in the presence of (and for the resolution of) duplicate individuals in overlapping catalogs. Overall, our results recommend the use of PIE v2 and CurvRank v2 for ID reconciliation in combination due to their complementary performance.
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 | 0.003 | 0.005 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 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".