Israel develops AI to detect pain in horses
Researchers in Israel have developed the SHIC-XE artificial intelligence system, which analyzes video and detects signs of pain in horses. The method was created by an international team led by Marcelo Feigelstein, head of the Artificial Intelligence Systems Engineering program at Tel-Hai University in Kiryat Shmona, The Jerusalem Post reports.
The system is designed not only to determine the presence of pain but also to explain which anatomical areas of the animal the model focuses on. According to Feigelstein, the team found that when making a decision, the model focuses, among other things, on the horse's ears and cheek muscles.
Stable video analysis
Traditional explainable artificial intelligence systems use two-dimensional attention maps that highlight important parts of an image. When working with video, such maps can change due to the movement of the animal or the camera, making them difficult to interpret and compare between frames.
To solve this problem, SHIC-XE projects the model's attention data onto a fixed three-dimensional representation of a horse's face. This makes it possible to maintain an anatomically consistent explanation when the animal moves, changes the position of its head, or appears in the frame from a different angle. The scientists said that the explanations generated by the system can be quantitatively compared with expert assessments.
More current news is available on the UA.News Telegram channel Telegram.
Potential applications
Professor Anna Zamanska and the Tech4Animals laboratory, Professor Ilan Shimshoni of the University of Haifa, Technion students Omer Bibi and Ofer Rosenbaum, as well as researchers from the universities of Bern, Milan, São Paulo and Newcastle participated in the study. The results were published in the International Journal of Computer Vision.
Feigelstein noted that animals often hide their suffering, making it difficult even for veterinarians and caregivers to recognize pain. Previously, his team also developed AI tools to recognize pain and emotions in cats, dogs, rabbits, sheep and cattle.
Currently, SHIC-XE identifies the presence of pain. In the future, the researchers plan to train the system to assess its intensity and possibly determine the likely origin of the pain. According to Feigelstein, the technology could potentially be used on farms, in veterinary hospitals, laboratories and during animal transport, but an application still needs to be created for practical use.
