Tinkoff Center for Artificial Intelligence Technologies developed and launched Russia’s first phone number scanner in a mobile app, which recognizes any phone number, both typed and handwritten.

Tinkoff press service reports that with its help customers can make a record-breaking fast money transfers by scanning any phone number in any form. Tinkoff developed its own handwriting recognition technology to create the scanner.

The scanner recognizes phone number digits based on neural networks and algorithms that have been trained on large data sets. Now the technology can recognize and read from any surface any phone number, printed or written in any form.

When pointing a smart camera at a phone number, the scanner analyzes the image, finds areas with numbers and combines them into lines, removing information that is not related to the phone number. Further, the identified lines are recognized using a neural network, from the received options, the search algorithm selects the most appropriate sequences of numbers related to the phone numbers of Russian operators.

The scanner processes one image instantly – in less than 0.3 seconds. If several phone numbers are displayed on the surface at the same time, the scanner reads them all and prompts the user to select the desired one.

“The strength of the Tinkoff team is that we, based on the latest technological trends, use the test and learn method, to create our own solutions to help users solve everyday tasks. When we started working on the scanner, we realized that the standard handwritten data sets are not suitable for our tasks – in real life, people write numbers differently.

Therefore, for the scanner to read handwritten phone numbers quickly and clearly, we developed our own generator of all kinds of handwriting, which made it possible to create number recognition technology,” Konstantin Markelov, Business Technology Vice President at Tinkoff, said.

The partner of Fintech section is