From video to behaviour: an LSTM-based approach for automated nest behaviour recognition in the wild
Bibliographic record
Abstract
ABSTRACT Studies of animal behaviour usually rely on direct observations or manual annotations of video recordings. However, such methods can be very time-consuming and error-prone, leading to sub-optimal sample sizes. Recent advances in deep-learning show great potential to overcome such limitations, nevertheless, most currently available behavioural recognition solutions remain focused on captivity settings. Here, we present a deployment-focused framework to guide researchers in building behavioural recognition systems from video data, using Long Short-Term Memory (LSTM) networks to classify behavioural sequences across consecutive frames. LSTMs allowed to: 1) monitor nest activity by detecting the birds’ presence and simultaneously classifying the type of trajectory: i.e., nest-chamber entrance or exit; and 2) identify the behaviour performed: building, aggression or sanitation. Using our framework, we outperformed human annotators when jointly considering error and speed. Model performance improved with challenging training instances, and remained robust even with modest sample sizes. LSTM also outperformed YOLO (“You Only Look Once”), highlighting the critical role of temporal sequence information in behavioural analysis. We demonstrate that our approach is replicable across three bird species and applicable to deployment videos, highlighting its value as a generalizable and transferable tool for long-term studies in the wild. DATA AVAILABILITY Scripts, models, and data required to reproduce this work are available on Zenodo (DOIs: 10.5281/zenodo.18681623 and 10.5281/zenodo.18695178).
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 | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| 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) | 0.003 | 0.001 |
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".