Horse welfare in competition assessed by AI

Can the well-being of a horse ridden in competition be verified using an artificial intelligence (AI) tool? A team of ethologists from the French National Institute for Agricultural, Food and Environmental Research (INRAE) has answered in the affirmative. Their machine learning program, "trained" using photographs that they annotated on various criteria, including those of British researcher Sue Dyson , is 90% accurate in detecting discomfort in a photo or video. "This work, at the proof of concept stage [demonstration of feasibility] , has shown that the expressions of the horse's head are statistically more discriminating [useful] than its postures in determining the animal's feelings," explains Léa Lansade, who led this study.

This research, funded by the French Equestrian Federation's Equiaction fund, was launched in response to growing societal interest in animal welfare, particularly during competitions. "This work is important for shedding light on this issue at a time when the use of horses in competition is now under scrutiny following the problems that occurred at the Tokyo Olympic Games . " adds Vincent Boureau, vice-president of the French Equine Veterinary Association, a leading authority on animal welfare issues. In 2021, images of a German rider excessively whipping her horse during the modern pentathlon events went viral, leading to the disappearance of horse riding in this discipline for the Los Angeles Olympics in 2028.

Building on these initial results, the INRAE ​​team wants to enrich the algorithm with data on facial expressions collected during previous work. "To obtain more qualitative information, we use the FACS method [facial action coding system ] adapted to horses. Each anatomical movement of the head is recorded as an "action unit," then statistically analyzed in a network. This allows us to decipher one by one the horses' facial expressions in relation to their positive or negative emotions," explains Léa Lansade.

You have 29.84% of this article left to read. The rest is reserved for subscribers.