Imagine a future where ranchers can check the health of a cow by simply looking at it’s face. CattleFever is a system that was developed by the University of Arkansas to make this futuristic concept a reality.
The tool was developed by the Artificial Intelligence and Computer Vision Lab. It uses thermal cameras and AI to measure the body temperature of cows. This is a first step towards automatic systems which could revolutionize the way farmers take care of their animals.
Trong Thang Pham was the project leader, and a PhD student from the University of Arkansas. He worked under Ngan Le’s guidance, an associate professor of computer science, electrical engineering, and mathematics. Le’s laboratory specializes in computer vision, robotics, and medical imaging.
They set out together to resolve a problem that has existed for a very long time: cattle temperature is currently measured by rectal measurement, which can be stressful for the animals, and labor intensive for ranchers. CattleFever is a noninvasive solution that can improve animal welfare, and detect diseases before they spread.
Edge AI can listen to silent cattle voices
Researchers needed data to train the system. Researchers needed data for dogs, cats and horses. The only dataset for cattle, CattleEyeView included RGB images for herd tracking.
The team created their own dataset in the Savoy Research Complex of the Arkansas Agricultural Experiment Station. Researchers recorded thousands of calves using short videos, thermal cameras and rectal thermometers.
Thermal image of an animal that is used to measure its temperature. Credit: University of Arkansas
Researchers then marked 13 landmarks on the face, including eyes, muzzles, mouths, and ears. CattleFace – RGBT is a dataset which links thermal and visible data. Now, the landmark-detection software can automatically detect a calf’s face as well as key features in both RGB and Thermal images.
Can AI accurately estimate the temperature of a cow by its facial expression?
The team found that the readings of the nose and eyes were the most similar to those taken by a thermometer. The system focused its attention on the thermal data of these regions using landmarks.
Climate change is increasing the risk of heat stress in livestock.
Random forest regression was the most accurate method of machine learning. It averages out results from multiple decision trees. CattleFever was able to estimate cows’ temperatures within one degree of the reading on a rectal temperature meter.
Currently, it works better when the cattle are facing the camera. Next, AI must be taught to identify cows in their natural environment, whether they are grazing or moving.
We need to get more pictures of the animals in their natural environment, like running, so we can capture how they move. Pham explains.
Researchers are invited to use the CattleFace dataset and build upon it.
We share it with everyone. “That’s what we do.” Pham stated.
CattleFever is a step towards precision livestock farming. AI and sensors will help ranchers to care for their animals in a more efficient and humane way. Ranchers can prevent outbreaks of fever and increase herd health by recognizing subtle changes in the cow’s facial expression.
In the future, by simply looking into a cow’s eyes, you could reveal more about its health, and it would help keep herds healthier and farmers to farm smarter.
Journal Reference
- Trong Thang Pham, Ethan Coffman, Beth Kegley, Jeremy G. Powell, Jiangchao Zhao, Ngan Le. CattleFever: An automated cattle fever estimation system. Smart Agricultural Technology. DOI: 10.1016/j.atech.2025.101434
