VideoMatrix tested its AI-based solution in Cherkizovo Group

2022: Testing an AI-based Vmx SILA solution to accurately record pigs

On May 23, 2022, VideoMatrix announced that testing of the Vmx SILA: LSI software from VideoMatrix had ended on the territory of the Cherkizovo Group pig farm. The solution is video analytics based on artificial intelligence and machine vision. The tested system automatically accounted for the number of pigs in real time when transferring animals to growing.

Photo: behance.net

Manual counting of pigs is traditionally practiced at Russian pig farms, but in conditions of current technologies it has a number of drawbacks - from accuracy of counting to risks associated with economic security. The underestimation of each animal carcass is critical for the enterprise. Automated accounting of animals makes it possible to reduce the time for counting pigs, to eliminate human errors in a routine labor process.

The basis of the already installed video equipments pig complex "Cherkizovo Group" without the use of foreign smart cameras was integrated domestic software from a Russian developer. The line Vmx is known in the market for high accuracy due to the hybrid combination neural networks and mathematical. During algorithms the last month of testing, more than 99% accuracy of counting was shown.

Case with Cherkizovo set a high level of requirements for our systems. It is one thing to detect defects on a sheet of the thinnest military steel on the PAC, another thing is to consider pigs in their chaotic unpredictable movement, distinguishing them from other objects and not losing any. Small individuals have become a special motivation for optimizing Vmx SILA: LSI: unlike large comrades-in-arms, their movement is much more intense, and a small carcass is periodically lost in the frame, finding itself larger under relatives. Therefore, we applied an additional development - proprietary tracking of the movement of each pig in the frame, including the "crowd management" tool. This is when temporarily lost-in-sight carcasses are taken into account by a smart algorithm in space relative to their neighbors, "- comments on the features of the technology, VideoMatrix CEO Farid Nigmatullin.

According to VideoMatrix, the automatic inventory of animals by artificial intelligence is also of interest to other large enterprises in the agro-industrial complex. A practical complex case proves that it is possible to achieve a product accuracy level of > 99.9%, which represents the opportunity to scale the digital tool to other pig complexes in the country with high requests for the reliability of the video analytics service.