Monday 12/January/2026 – 04:37 AM
Chinese scientists have succeeded in developing a lightweight model to recognize cow behavior based on videos captured by a four-legged robot working in grassy pastures, in a step that contributes to improving the efficiency of feeding and managing livestock herds, according to what was reported by the Chinese news agency Xinhua.
The Agricultural Information Institute of the Chinese Academy of Agricultural Sciences said it had proposed a new model called Masem-Yulu, noting that the research results were published in the Journal of Computers and Electronics in Agriculture.
The researchers explained that accurate and rapid identification of typical cow behaviors is an essential element in diagnosing diseases, monitoring the reproductive cycle, predicting calving, and assessing the health status of livestock.
The MASM-YOLO model is distinguished by its ability to accurately detect several behaviors in complex environmental conditions, with the possibility of implementing it in real time on board mobile robots, which makes it suitable for field use.
Adaptive analysis and distribution
By integrating a multi-scale focusing and extraction network, an adaptive analysis and distribution head, and other advanced technologies, the model successfully addresses key challenges such as lighting changes, motion blur, and occlusion within livestock groups.
MASM-YOLO enables rapid recognition of six typical beef cattle behaviors, including feeding, resting, locomotion and licking, achieving an ideal balance between recognition accuracy and computing efficiency.
The researchers emphasized that this model represents important technical support for the development of smart grazing robots, and paves the way for enhancing digital transformation in the livestock sector.
