AI Unlocks Horses' Body Language for Superior Veterinary Care

Scientists are leveraging artificial intelligence to close the communication divide between horses and humans. By integrating 3D motion capture technology with advanced machine learning algorithms, they have developed a novel modeling system. This innovative tool will provide veterinarians with an enhanced visual aid to decode equine body language—essential for identifying both physical ailments and behavioral issues.

According to recent findings from KTH Royal Institute of Technology and Swedish University of Agricultural Sciences (SLU), the system has been developed to recreate precise 3D movements of horses seen in video footage. This technology utilizes an AI-driven parametric model that captures both posture and form of the animals.

The model is accurate enough to allow a veterinarian, for instance, to identify significant alterations that might otherwise go unnoticed or misunderstood during an evaluation, like variations in a horse’s stance or their body mass. The research demonstrates this effectively. published on the arXiv preprint server.

The system known as Dessie utilizes disentangled learning, allowing it to isolate various crucial elements within an image and assist the AI in not getting confused by background elements or lighting variations, according to Hedvig Kjellström, a professor of computer vision and machine learning at KTH.

Kjellström states that Dessie represents the initial instance of disentangled learning in three-dimensional motion models not created by humans.

Elin Hernlund, an Associate Professor specializing in biomechanics at SLU and a clinician for equine orthopedics, suggests that Dessie could lead to increased precision in observing and interpreting horses’ movements. Consequently, this might allow for earlier and more accurate interventions compared to current methods. Essentially, it provides crucial data directly “from the source.”

Hernlund explains that horses possess strength yet remain vulnerable creatures, communicating their state through body language. She adds that observing their movement allows them to detect signs of discomfort, such as an altered gait.

Hernlund states, “We have developed a digital voice aimed at helping these creatures overcome the communication gap between animals and humans. It’s designed to convey their feelings to us.” He adds, “This technology represents the most intelligent and high-resolution method for capturing digital data from horses—including subtle cues from their facial expressions—that can provide significant insights into their state.”

The researchers are continuing to train Dessie using images from an expanded range of horse breeds and sizes. This will allow them to connect genetic information with physical traits, thereby enhancing their comprehension of the fundamental biological architecture of these creatures.

"To accomplish this, we're requesting breeders to submit photos of their breeds to encompass as much diversity as possible," Hernlund states.

More information: Ci Li and colleagues, Dessie: Disentanglement for Articulated 3D Horse Shape and Pose Estimation from Images, arXiv (2024). DOI: 10.48550/arxiv.2410.03438

Furnished by KTH Royal Institute of Technology

This tale was initially released on . Subscribe to our newsletter For the most recent science and technology news updates.