Artificial intelligence is finding its way into livestock farming, with Indian researchers developing a new AI-based system that can continuously watch and track the behaviour of Mithun in a natural farm environment.
Researchers at the ICAR-National Research Centre on Mithun (ICAR-NRC on Mithun), Nagaland, have developed the technology to help farmers and livestock managers monitor animals without continuous physical observation. The research has been published in Engineering Research Express.
Mithun (Bos frontalis), often called the “Cattle of the Hills”, has strong social, cultural and economic importance for tribal communities in Northeast India. The animal is closely connected with rural livelihoods and food security in the region.
The newly developed AI system can automatically identify four important behaviours of Mithun — feeding, standing, lying and mounting. These everyday activities can reveal useful information about an animal’s health, comfort, nutrition and physical condition.
For instance, an unusual change in feeding or lying behaviour could indicate a possible health or welfare problem. Tracking mounting behaviour can also help farmers understand reproductive activity and support better oestrus and breeding management.
To develop the technology, researchers installed 12 high-definition CCTV cameras across two sheds at the ICAR-NRC on Mithun farm in Nagaland. The cameras continuously monitored the animals during the day and night. Infrared technology was used for monitoring under low-light conditions.
Researchers then created a dataset of 3,000 manually labelled images covering the four selected behaviours.
The system uses two technologies — the YOLOv8n object-detection model and DeepSORT tracking technology. In simple terms, one identifies what the animal is doing, while the other follows individual animals across video footage. This allows the behaviour of each Mithun to be monitored in real time.
The results showed a high level of accuracy. The YOLOv8n model achieved a mean average precision of 99.5 per cent and a recall of 99.6 per cent. The system was able to process around 31 video frames per second using an NVIDIA RTX 3060 GPU.
Importantly, researchers also tested the system under real farm conditions where monitoring is not always easy. These included animals partially blocking each other, wet and uneven ground, shadows, background clutter, motion blur and nighttime footage.
The technology could reduce the need for workers to continuously watch animals, a process that is both time-consuming and difficult to maintain throughout the day and night. Instead, AI-based monitoring could provide livestock managers with regular information and help them notice behavioural changes earlier.
However, researchers say more testing is required before the technology can be widely used. The present system has been tested at only one farm and currently tracks four behaviours.
Future research could expand the technology to identify aggression, grooming and disease-related inactivity. Researchers are also exploring larger datasets and systems that can work across different farms, seasons and environments.
The research brings together artificial intelligence, computer vision and livestock science, showing how digital technology could support better animal health, welfare, breeding and farm management.
The study involved scientists from ICAR-NRC on Mithun along with researchers from NIT Nagaland, Nagaland University and CHRIST (Deemed to be University).