Dairy DigiD: a keypoint-based deep learning system for classifying dairy cattle by physiological and reproductive status
Bibliographic record
Abstract
Precision livestock farming increasingly relies on non-invasive, high-fidelity systems capable of monitoring cattle with minimal disruption to behavior or welfare. Conventional identification methods, such as ear tags and wearable sensors, often compromise animal comfort and produce inconsistent data under real-world farm conditions. This study introduces Dairy DigiD, a deep learning-based biometric classification framework that categorizes dairy cattle into four physiologically defineda groups-young, mature milking, pregnant, and dry cows-using high-resolution facial images. The system combines two complementary approaches: a DenseNet121 model for full-image classification, offering global visual context, and Detectron2 for fine-grained facial analysis. Dairy DigiD leverages Detectron2's multi-task architecture, using instance segmentation and keypoint detection across 30 anatomical landmarks (eyes, ears, muzzle) to refine facial localization and improve classification robustness. While DenseNet121 delivered strong baseline performance, its sensitivity to background noise limited generalizability. In contrast, Detectron2 demonstrated superior adaptability in uncontrolled farm environments, achieving classification accuracies between 93 and 98%. Its keypoint-driven strategy enabled robust feature localization and resilience to occlusions, lighting variations, and heterogeneous backgrounds. Cross-validation and perturbation-based explainability confirmed that biologically salient features guided classification, enhancing model transparency. By integrating animal-centric design with scalable AI, Dairy DigiD represents a significant advancement in automated livestock monitoring-offering an ethical, accurate, and practical alternative to traditional identification methods. The approach sets a precedent for responsible, data-driven decision-making in precision dairy management.
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 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 | 0.001 | 0.001 |
| 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 | 0.000 |
| 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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, 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".