MODERN TEXT RECOGNITION MODELS AND METHODS: TECHNOLOGIES, CHALLENGES, AND PERSPECTIVES

Abstract The article focuses on modern methods and models of Optical Character Recognition (OCR). It reviews key approaches, including classical algorithms, Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), their combinations (CRNN), and transformers. A comparative analysis of their advantages and limitations is presented, and an optimal approach is proposed for processing textual data in real-world conditions. The practical section demonstrates the efficiency of the CRNN method in text recognition tasks considering factors such as low image quality and complex backgrounds.