Photo by Y. Ivanko / mos.ru
Russian scientists have developed a neural network capable of recognizing Russian handwritten texts with 99% accuracy, the website of the Ministry of Education and Science of the Russian Federation reports. The work was carried out by experts from the Siberian Federal University (SibFU) and the Electrotechnical University in St. Petersburg.
The network recognizes the letters of the Russian alphabet, written by hand. The application created for the neural network does not require an Internet connection and is reliably protected from data leaks, the user can work safely and confidentially.
One of the co-authors of the study, SibFU student Andrei Levkov, noted that this is the first such development in the world. According to him, the task of converting handwritten text into a digital version is very relevant.
The neural network was trained using data from the CoMNIST repository, which contains sample texts in Cyrillic and Latin. The artificial intelligence processed more than 13,000 images with uppercase, printed and cursive letters. About 85% of the photos were used for training, the remaining 15% were used to test the accumulated ''knowledge''.
Now scientists intend to train the neural network in various styles of writing and "reading" words and sentences.
Russkiy Mir
