Researchers from KTH Royal Institute of Technology and Swedish University of Agricultural Science have developed a platform called DESSIE that can reconstruct the 3D motion of horses from videos using an AI-based parametric model. This model allows for precise analysis of a horse’s posture and body weight, enabling veterinarians to spot changes that may be overlooked during an examination.
DESSIE employs disentangled learning, separating important factors in an image to avoid confusion with background details or lighting conditions. This marks the first example of disentangled learning in non-human 3D motion models, according to Professor Hedvig Kjellström. Elin Hernlund, an Associate Professor in biomechanics at SLU, believes that DESSIE will enable greater accuracy in observing and interpreting horses’ movements, leading to earlier and more precise intervention.
The goal of DESSIE is to provide a digital voice for horses, allowing them to communicate their feelings and needs to humans. By analyzing the horse’s gait and body language, veterinarians can detect signs of pain or discomfort. The research team plans to further train DESSIE with images of different horse breeds and sizes to better understand the biological structure of animals and link genetics to phenotypes.
