Russian Scientists Build New Database to Train AI to Estimate Cow Weight at a Glance
A detailed digital “body map” of a cow is built around 38 anatomical reference points.
Russian researchers at the Russian State Agrarian University – Timiryazev Moscow Agricultural Academy have registered a first-of-its-kind database designed to train artificial intelligence systems for use in livestock farming. The dataset allows computers to accurately “see” and analyze a cow’s posture using ordinary images and video, opening the door to smart systems that monitor animal health and productivity without direct contact.
The detailed digital body map of a cow is based on 38 anatomical points. By comparison, many existing systems rely on just 17. This dense skeletal annotation enables algorithms to reconstruct body geometry more precisely, even when an animal is moving, partially obscured, or captured from challenging angles. The AI learns to infer missing points when they are not visible.
Stress-Free Weighing
The database contains more than 12,000 carefully annotated images. That scale ensures models trained on the data can perform reliably in real farm conditions.
As a practical application, the technology could be used to develop a system for precise, contactless weighing. By analyzing the spatial arrangement of key points, AI can calculate an animal’s weight without stressful handling procedures. Continuous skeletal analysis can also be used to monitor gait and detect early signs of lameness or other health issues.
The development makes a significant contribution to the digital transformation of agriculture. The database will also serve as a valuable educational resource, giving students and early-career researchers access to unique, hands-on training material.
