The intensity and productivity of pig production increased significantly during the past 20 years. However, jointly with the increase of the number of born piglets, also the perinatal piglet mortality increased. Currently, on average about 25% of the born piglets dies before the age of 10 weeks, which is equivalent to 8.9 deceased piglets per sow per year. Typical target values for piglet mortality are <7% stillborn, <12% deceased during the suckling period and <2% in the nursery barns. However, in practice, some farms report perinatal piglet mortality rates up to 30%. The goal of this project is to reduce perinatal piglet mortality related to non-infectious causes and risk factors such as a low birth weight, a long inter-piglet birth interval, piglet crushing by sows, aggressive behaviour of sows, etc. To achieve this goal, we make use of innovative tools and monitoring methods. One of these innovative tools is behaviour monitoring of sows and piglets in the farrowing units by means of computer vision. By developing computer vision models for animal detection, behaviour classification, animal tracking and pen detection, we aim to get a profound insight into the behaviour characteristics of sows and piglets in the farrowing units. Moreover, we aim to elucidate inter-animal variance in behaviour characteristics which can be related to perinatal piglet mortality. Preliminary results show that sow behaviour before farrowing is highly predictive for sow behaviour after farrowing. Sows which are more nervous before farrowing, will remain more nervous during the complete lactation period. However, restlessness alone does not seem to have a strong correlation with perinatal piglet mortality and therefore further research and analysis are required.

Source: Annual AgData Use Case Summit 2026 booklet.