Why sensor-free?
Collars, pedometers and boluses require one device per animal. Batteries die, devices get lost, broken units are replaced, and cost scales linearly with the herd. With a camera-based approach, cost depends on the number of monitored points rather than the number of animals: one camera at a passage sees every animal that walks through it.
The models were trained on our own farms
The dairy facility in Konya belonging to founder Safa Batuhan Aslantaş’s family became the system’s first site. The models were trained not on an off-the-shelf dataset but on footage captured in real barn conditions: changing light, mud, crowding and genuine animal behaviour. Development involved more than 9,000 dairy cows and over 22 veterinarians.
From vision to decision
The animal is first detected and identified in the camera feed. Gait, posture and body outline are then analysed. The resulting scores are compared against the animal’s own history — because what matters is the trend, not a single day’s score. Any animal past the threshold enters the watchlist and, where warranted, triggers a notification.
Where does the data live?
Recordings and generated scores are stored in the cloud, so an animal’s trend can be reviewed months back. The footage belongs to the farm; it is not shared with third parties other than the infrastructure provider that hosts it, and it is not used for model training without the farm’s separate written consent.