Machine Learning Applications in Pig Welfare

Machine learning has become integral to modern pig welfare management by enabling continuous, objective monitoring of animal health and behaviour. Through the integration of sensors—including depth and RGB cameras, microphones and accelerometers—with advanced algorithms, subtle deviations in posture, activity and feeding can be captured and interpreted in real time. These approaches facilitate early detection of disease, stress or aggression, reducing reliance on manual observation, optimising interventions and promoting sustainable husbandry. By shifting from group-level assessment to individual pig surveillance, precision livestock farming enhances diagnostic accuracy, refines welfare outcomes and supports global efforts to mitigate antimicrobial use and improve production efficiency.

Research from Nature Portfolio

A foundational 2017 study introduced an automated system employing depth video cameras and machine learning to track pig movement in three dimensions. By extracting standing, feeding, drinking and locomotor activities from 3D trajectories, the system reliably detected behavioural changes induced by a controlled challenge, demonstrating its utility as an early-warning tool in commercial barns. Subsequent work has focused on broadening the range of detectable behaviours, improving robustness across varied facility layouts and integrating continuous alerts for stockpersons, thereby laying the groundwork for scalable real-time welfare surveillance systems.

Machine Learning Applications in Pig Welfare publication trend

The graph below shows the total number of articles in machine learning applications in pig welfare across all publications each year (not limited to Nature Index journals).

Technical terms

Convolutional neural network (CNN): A deep learning architecture that applies hierarchical filters to extract spatial patterns from images.

XGBoost: A scalable gradient-boosted decision tree algorithm widely used for classification and regression.

Average Precision (AP): The area under the precision–recall curve, assessing detection accuracy across confidence thresholds.

Mean Average Precision (mAP): The average of AP values over multiple classes or tasks, indicating overall detection performance.

Area under the curve (AUC): The integral of the receiver operating characteristic curve, reflecting a model’s ability to discriminate between classes.

References

1. Disease detection in pigs based on feeding behaviour traits using machine learning. Biosystems Engineering (2023).
2. Automated tracking to measure behavioural changes in pigs for health and welfare monitoring. Scientific Reports (2017).
3. Towards on-farm pig face recognition using convolutional neural networks. Computers in Industry (2018).
4. Automatically detecting pig position and posture by 2D camera imaging and deep learning. Computers and Electronics in Agriculture (2020).