By Sehrish Malik
Piglet mortality remains a major challenge in pig production, with over 18% of piglets lost before weaning (Ingris 2023). One key factor influencing piglet survival is the sow’s behaviour before and after farrowing. However, observing these behaviours manually is time-consuming, subjective, and often disruptive to the animals.
In our study, we developed computer vision algorithms that automatically detect sow postures and nursing behaviour. We classify postures into five categories: standing, kneeling, sitting, sternal lying, and lateral lying. These posture patterns help us estimate activity levels, such as nest-building behaviour, and detect restlessness before farrowing.
Early results show that our algorithms reliably detect posture changes, enabling analysis of motion intensity and frequency, as well as nursing-like episodes to gain early insights into nursing patterns. These initial results offer a first glimpse into individual behavioural differences, indicating that AI can help us monitor sow behaviour continuously and non-invasively. Understanding individual sow behaviour will enable us to identify maternal traits that support piglet survival and guide breeding and management strategies.By
