Guwahati: Cameras and artificial intelligence are being trained to keep a round-the-clock watch on Mithun (Bos frontalis), the iconic “Cattle of the Hills”, in a research effort that could change how the animals are monitored, bred and cared for.
Researchers at the ICAR-National Research Centre on Mithun have developed an AI-powered system that can detect and track Mithun behaviour in real time without disturbing the animals.
The system uses 12 high-definition CCTV cameras installed across two sheds at the ICAR-NRC on Mithun farm. With infrared coverage, the cameras monitor the animals day and night.
Researchers trained the system using 3,000 manually annotated images covering four behaviours—feeding, standing, lying and mounting.
The technology combines YOLOv8n, an AI model used for object and behaviour detection, with DeepSORT, which tracks individual animals across video frames and assigns them persistent identities.
The results were striking. The detection model achieved 99.5% mean average precision ([email protected]) and 99.6% recall, while processing footage at about 31 frames per second on an NVIDIA RTX 3060 GPU.
More importantly, the system was tested in the messy conditions of a working farm, including shadows, motion blur, wet and uneven ground, background clutter, partial obstruction and nighttime infrared footage.
Why it matters
For farmers, animal behaviour can be an early warning system. A change in feeding, standing or lying patterns may indicate problems with health, nutrition or comfort. Mounting behaviour, meanwhile, can provide useful clues for reproductive and oestrus management.
At present, such observations largely depend on people watching animals manually—something that becomes difficult to sustain around the clock, particularly at night.
An automated system could allow farmers and livestock managers to track these behavioural changes continuously and use the data to make decisions about animal health and breeding.
Mithun has a particularly important place in Northeast India, where it carries social and cultural significance for tribal communities and contributes to livelihoods and food security.
But the researchers caution that the technology is still at an early stage. It has so far been tested on a single farm and needs validation across different farms, seasons, regions, stocking densities and camera configurations. Severe obstruction of animals can also affect its performance.
The next phase could make the system considerably smarter, with researchers looking at behaviours such as aggression, grooming and disease-related inactivity. They are also exploring temporal AI models, edge-device deployment and larger datasets covering different farms and seasons.
The study, published in Engineering Research Express, was carried out by researchers from ICAR-NRC on Mithun in collaboration with NIT Nagaland, Nagaland University and CHRIST (Deemed to be University).
If scaled successfully, the technology could take Mithun monitoring from occasional human observation to a continuous, data-driven system—giving farmers another set of eyes on their herds, day and night.