TrueType Transformer: Character and Font Style Recognition in Outline Format

Abstract

We propose TrueType Transformer (T 3 ), which can perform character and font style recognition in an outline format. The outline format, such as TrueType, represents each character as a sequence of control points of stroke contours and is frequently used in born-digital documents. T 3 is organized by a deep neural network, so-called Transformer. Transformer is originally proposed for sequential data, such as text, and therefore appropriate for handling the outline data. In other words, T 3 directly accepts the outline data without converting it into a bitmap image. Consequently, T 3 realizes a resolution-independent classification. Moreover, since the locations of the control points represent the fine and local structures of the font style, T 3 is suitable for font style classification, where such structures are very important. In this paper, we experimentally show the applicability of T 3 in character and font style recognition tasks, while observing how the individual control points contribute to classification results.

Strokes classification of handwritten text based on the Frechet distance

“…Задача перевода изображений рукописного или печатного текста в текстовые данные (OCR) популярна уже давно. Точность для распознавания печатных документов по чётким изображениям в настоящее время превышает 99% [1,2]. Точность для распознавания документов с рукописным текстом составляет 80%-90%, и как правило она достигается с применением нейронных сетей.…”