Handwritten Text Recognition HTR is the process of extracting handwritten text from an image and converting it into a digital form for machine operation. The existing approaches to solve the HTR are based on Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), which utilize the Connectionist Temporal Classification (CTC) objective function, and approaches based on attention models (Seq2Seq). In this article, we provide extensive comparison of the current Deep Learning approaches for the task of HTR. Also, we outline the current problems that limits the effectiveness of these approaches.

Мохаммед С. (науч. рук. Тесля Н.Н.) Deep Learning for Handwriting Text Recognition: Existing Approaches and Challenges // Сборник тезисов докладов конгресса молодых ученых. Электронное издание. – СПб: Университет ИТМО, [2022]. URL: https://kmu.itmo.ru/digests/article/9263