A neural network has been created in Russia to estimate the weight and predict the growth of meat in animals

Today, in industrial animal husbandry, the assessment of the exterior and productivity of livestock is carried out by an expert in manual mode, which makes it subjective. Traditional livestock weighing is stressful for the animal, which can lead to weight loss. The method proposed by scientists eliminates both of these risks.

The system, using computer vision technology, allows you to measure the weight and determine the morphological characteristics of the animal, and in the future, predict the change in weight.

“Without touching the cow, we determine its body weight and assign a category of MEAT. For example, it's no secret that gourmets highly value marbled beef . You can predict how much such a delicacy will turn out when the calf grows up. The monitoring that we carry out allows us to make the right decision about the optimal time for slaughtering livestock, which allows us to reduce the cost of raising livestock,” said the author of the methodology, HEAD of the Department of Computer Security and Applied Algebra of the ChelGU Faculty of Mathematics Alexei Ruchai.

Now the scientific team is engaged in replenishing the database, the size of which determines the efficiency and accuracy of the system. To do this, it is necessary to constantly conduct experiments on non-contact weighing of different animals. So far, the error of non-contact measurement is 5-10%, which is comparable to conventional weighing.

Initially, the experiment was built on the basis of three chambers. The researchers needed to calibrate the cameras and fix the animal in a motionless state. Now experts are working on a technique for measuring with a single camera, past which the animal passes. The goal of the scientific team is to create a simple and versatile automated non-contact weighing system that anyone can set up without special training.