Specialists from the Russian company Smart Engines have deciphered crossed-out fragments of Alexander Pushkin's draft manuscripts using the "Da Vinci" neural network, according to Dmitry Nikolaev, Head of the "Visual Systems" Department at the Federal Research Center for Computer Science and Control of the Russian Academy of Sciences.
The artificial intelligence was trained to recognize Pushkin's pen movements in non-crossed-out words and use this data to restore lost elements of the text. The results of the work confirmed that AI can contribute to the study of the great poet's creative process.
According to Vladimir Arlazorov, CEO of Smart Engines, the method of removing strikeouts using neural networks may be useful not only in the study of Pushkin's works, but also in working with archival records of other authors and historical documents.
The technology automatically determines the geometry of the document and recognizes data regardless of its location in the frame. For example, recognizing Russian passports using a smartphone camera takes only 0.15 seconds, and the company's server solutions can process up to 55 passports per second without using a GPU.
Earlier, www1.ru reported that scientists trained a neural network on sperm whale "songs."
Read more on the topic:
Russia wants to train AI to create cancer vaccines
A neural network for calculating holograms — 3D-CGH-Net — has been created in Moscow
Scientists at MTUCI will teach a neural network to read damaged barcodes
