The Digital Economy Development Institute (DEDI LLC), a resident of the Skolkovo Foundation’s IT Cluster, announces the new version of the SmartLook Retail image recognition system, which has increased the accuracy of price tags on store shelves.

DEDI’s researchers and developers have created a new algorithm using several neural networks and modified computer vision algorithms. The algorithm showed an accuracy of 98.6% in recognizing the main price (rubles, dollars, etc) on price tags when tested in a number of retail chains.

When determining prices with small or poorly printed figures (pennies, cents, etc), the accuracy is now 90 +%. The peculiarity of the new solution is that the algorithm does not reject the entire price tag due to poor print quality or low-resolution photos, but provides maximum data completeness - even if the pennies are visually indistinguishable, rubles will be recognized with high accuracy.

SmartLook Retail can be used in several versions - as part of the OPTIMUM automated mobile trading system and as an independent cross-platform mobile application running Android or iOS. In addition, third-party mobile trade automation (SFA) developers can quickly embed SkyNet into their applications.

Khuseyn Az-zari, DEDI’s General Director notes, “Thanks to optimized algorithms for working in a wide segment of mobile devices and unique machine learning technologies, the analysis of one photo with a large number of recognizable goods takes less than 1 second. The system works directly on a mobile device without accessing servers on the Internet, which allows you to instantly make recommendations for adjusting or replacing price tags, replacing advertising materials and correcting the display of goods directly to store employees or authorized merchandisers. In comparison with similar systems, a large amount of productive time is saved, there is no dependence on the Internet, costs and load on servers and other infrastructure are reduced.“