Animal behaviorists from Scotland’s Rural College in Edinburgh (SRUC) are using 3D facial recognition technology to interpret the highly expressive faces of pigs. Working with machine vision and machine learning experts at the University of the West of England (UWE Bristol), the team’s aim is to develop a tool that farmers could use to monitor individual animal faces and receive alerts about any health or welfare problems within their livestock, giving farmers the ability to improve the well-being of their animals and prevent any impact on production. A standard photo of a sow (left). A digitized image of the same sow (right) is analyzed using state-of-the-art 3D technology to help researchers and animal behaviorists discern the meaning of the animal’s facial expressions. Courtesy of SRUC. “It is important to properly validate the technology before implementing it on-farm, but it has the potential to develop a truly animal-centric welfare assessment technique that looks at individuals and could tell us something about the importance each individual animal places on their own particular experiences,” Emma Baxter, senior researcher at SRUC, told Photonics Media. “Individualizing the information could allow for targeted treatment protocols and allows insight into both short-term emotional reactions and long-term individual ‘moods’ of animals under our care,” she said. The study, funded by the Biotechnology and Biological Sciences Research Council, provides evidence that pigs signal their intentions through their facial expressions, and that these expressions also provide clues regarding their emotional states of being. Sows that experience lameness, for example, show with their faces when they are in pain and when they feel relief. Pigs exhibiting aggression will signal their intentions prior to attacking other pigs. Using 3D facial recognition technology, animal behaviorists are working to detect the different emotional states in pigs. Scientists at SRUC’s Pig Research Centre in Midlothian, Scotland, are using the latest technology to capture 2D and 3D images of the breeding sow population under various typical commercial situations to see the diverse emotional states they exhibit while under different conditions. These images are sent to UWE Bristol’s Centre for Machine Vision, where they’re analyzed and used to help develop machine learning techniques that will allow farmers to automatically recognize different emotions conveyed through their pigs’ facial expressions. In a previous study conducted by the team, simple motion detection cameras captured images of the pigs at the research center. The images were hand-identified and then sent to UWE to be run through adapted human facial-recognition models using convolutional neural networks. This led to the development of highly successful, 97% accurate deep learning algorithms. “We consider the facial recognition work as proof of concept and we are further optimizing the model to ensure it manages detection over time — that is, to deal with the fact that pigs age relatively fast,” Baxter said, adding that the team is taking an integrated approach to the research, focusing on aging to guarantee that the technology will be able to recognize an individual animal over time. Once the techniques are validated for pigs, the team at UWE Bristol will develop the cost-efficient technology in commercial partnerships with JSR Genetics Ltd., Garth Pig Practice, and AgsenZe. The study could be adapted and translated to validating the states of being of other species of farm animals, such as cattle, sheep, and goats, which are also highly expressive.