Advances in deep learning-driven photo identification and meta analysis of cetaceans in large data repositories
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Photo-identification of cetaceans remains a labor-intensive task, requiring expert annotation of long-tailed image datasets in which most individuals are rarely encountered. We present a scalable, end-to-end framework that automates this process using lightweight deep learning models optimized for resource-constrained environments. Our modular pipeline integrates state-of-the-art detection (YOLOv8-small), individual identification via metric learning (EfficientNet-B0 with a contrastive head), and auxiliary modules for image quality scoring, side classification, and identifiability prediction. Unlike previous approaches limited to single-species applications or high-resource settings, our framework generalizes across five cetacean populations with diverse visual characteristics. We achieve top-1 identification accuracies of 0.92 for Bigg's killer whales ( Orcinus orca rectipinnus ), 0.96 for Southern resident killer whales ( Orcinus orca ater ), 0.96 for Lahille's bottlenose dolphins ( Tursiops truncatus gephyreus ), 0.82 for common minke whales ( Balaenoptera acutorostrata scammoni ), and 0.85 for humpback whales ( Megaptera novaeangliae ), yielding a cross-species accuracy of 0.90. To support image triage in large datasets, we include a quality scoring module that predicts image utility using learned embedding features. This module achieves an R 2 of 0.799, enabling intelligent prioritization of data. Runtime evaluations show processing speeds of 1.6–3.2 images/s on CPU and 9.6–23.3 FPS with GPU acceleration, making it suitable for archival and real-time applications. We also evaluate the impact of demographic metadata (age, sex) on identification performance and provide practical recommendations for future dataset design. The system is available via a web interface designed to support real-world conservation workflows with minimal computational overhead.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 it