So far, studies have demonstrated 97% accuracy in identifying individual pigs using only the face, senior researcher in animal behaviour and welfare at SRUC, Dr Emma Baxter, told GlobalMeatNews. “We consider [this success rate] proof of concept and we are further optimising the model to ensure it manages detection over time,” she said.
Researchers first positioned simple motion detection cameras behind a water drinker at SRUC pigpens at its Pig Research Centre in Penicuik, Midlothian. “This was completely non-invasive and meant that animals could have multiple images collected without any disruption,” Baxter said.
A team at the University of the West of England’s (UWE’s) Centre for Machine Vision then developed deep learning algorithms using convolutional neural networks to scan images and focus on distinguishing facial features of individual pigs.
Researchers are now working to identify pigs’ facial expressions and detect changes that could indicate their emotional state – a potentially valuable welfare management tool. They do so by capturing 3D and 2D facial images of the breeding sow population under various biological events likely to result in different emotional states. Using these images, UWE’s Centre for Machine Vision is further developing state-of-the-art machine learning techniques to automatically identify different emotions conveyed by different facial expressions.
The team wants to develop the technology for on-farm use with commercial partners, in which individual sows in large herds would be monitored continuously.
If it can be scaled-up, changes in farm infrastructure would probably be required, she explained. For example, farms would need to be fully connected to uninterrupted power sources and reliable internet capabilities, while detection cameras would need to be installed at pig height, but out of their reach.
While the tool could not and should not replace good personal husbandry, Baxter said its ability to monitor animal health and wellbeing could facilitate timely and targeted intervention, especially in larger farm operations.
Dr Zoe Davies, chief executive of Britain’s National Pig Association, agreed that farming experience would still be critically important: “Though these tools could help to support that, there is so much more to an animal’s state than just its face,” she said.
She stressed how farmers could make comprehensive assessments, regularly assessing key indicators such as the physical farm environment, lameness, and past and present injuries. “In the future, these tools might be an additional resource and help put these outcomes into context, but nothing will replace the stockperson,” she concluded.