RUS  ENG
Полная версия
ЖУРНАЛЫ // Компьютерная оптика // Архив

Компьютерная оптика, 2022, том 46, выпуск 3, страницы 455–464 (Mi co1034)

Эта публикация цитируется в 3 статьях

INTERNATIONAL CONFERENCE ON MACHINE VISION

Handwritten text generation and strikethrough characters augmentation

A. V. Shonenkova, D. K. Karachevb, M. Yu. Novopoltseva, M. S. Potaninac, D. V. Dimitrovad, A. V. Chertokae

a Sber AI, Moscow, Russia
b OCRV, Moscow, Russia
c Moscow Institute of Physics and Technology (State University), Dolgoprudny, Moscow region
d Lomonosov Moscow State University
e AIRI, Moscow, Russia, Nizhny Susalny lane, 5, p. 19

Аннотация: We introduce two data augmentation techniques, which, used with a Resnet – BiLSTM – CTC network, significantly reduce Word Error Rate and Character Error Rate beyond best-reported results on handwriting text recognition tasks. We apply a novel augmentation that simulates strike-through text (HandWritten Blots) and a handwritten text generation method based on printed text (StackMix), which proved to be very effective in handwriting text recognition tasks. StackMix uses weakly-supervised framework to get character boundaries. Because these data augmentation techniques are independent of the network used, they could also be applied to enhance the performance of other networks and approaches to handwriting text recognition. Extensive experiments on ten handwritten text datasets show that HandWritten Blots augmentation and StackMix significantly improve the quality of handwriting text recognition models.

Ключевые слова: data augmentation, handwritten text recognition, strikethrough text, computer vision, StackMix, handwritten blots

Язык публикации: английский

DOI: 10.18287/2412-6179-CO-1049



© МИАН, 2024