Artificial Intelligence could help equine gait assessment
University of Florida scientists want to assess livestock mobility faster and more accurately, ultimately helping farm animal health and production.
To do so, they’ll use artificial intelligence (AI) to analyze high-definition video of the animals as they move.
Samantha Brooks, a UF/IFAS geneticist and associate professor of equine physiology – along with other UF researchers — have been awarded a $49,713 grant from the Agricultural Genome to Phenome Initiative (AG2PI) for this research.
The team will combine machine learning with gait analyzes to speed their assessment of livestock mobility.
Brooks cites an example of how this technology can help: In horses, one veterinarian can do a basic lameness exam in about 15 minutes.
“Our long-term goal is to build an automated pipeline that could produce results nearly in real-time, just seconds after the animal passes by the camera,” Brooks said. “This pilot project is a first step toward that goal.”
Brooks and her colleagues work primarily with horses because they’re an excellent model for locomotion and because scientists can gather a lot of data quickly.
She and her lab already are working with about 2,000 video clips of horses in motion. Brooks credits the hard work of graduate student Madelyn Smythe and the generosity of hundreds of central Florida horse owners for the video.
“The large library of video will enable construction of accurate models to track the animals’ movement in the video frame,” Brooks said. “Although we’ve started with the horse, what we learn here will translate to similar models for other four-legged farm animals.”
For this project, they’ll also build AI models to analyze video of cattle, swine and small ruminants.
As they review the data, researchers will look at such horse traits as stance time, stride length and limb extension. In cattle and swine, scientists are more interested in asymmetry and postures that indicate pain for abnormal function in one or more limbs.
Brooks said she wants to help other scientists and owners of farm animals because AI, while helpful, is not always intuitive.
“Artificial intelligence approaches can accelerate our ability to measure complex locomotor traits in livestock, with better accuracy than the human eye,” Brooks said. “Yet, AI tools often are not biologist-friendly, nor are they ready for challenging, on-farm applications.”
“To cope with these issues, we hope to adapt and assemble existing AI methodologies into an analysis package accessible to scientists of diverse backgrounds and deployable in a variety of livestock management settings,” she said.
For instance, the technology could detect lameness in livestock as they pass by a camera each day. Envision dairy cattle coming into the parlor, for example — alerting the farmer to potentially serious health issues early on, and with less effort from farm staff.
Funded by the US Department of Agriculture’s National Institute of Food and Agriculture, AG2PI is a three-year project ending in 2023.
The goal of AG2PI is to connect crop and livestock scientists to each other and to those working in data science, statistics, engineering and social sciences to identify shared problems and collaborate on solutions.