Handwriting Recognition and Scene Text Recognition
|Handwriting Recognition and Scene Text Recognition
|Elisa Hope Barney Smith <email@example.com>
|Luleå tekniska universitet
|2024-01-30 – 2024-08-01
This proposal addresses the task of text recognition around two projects: a first dealing with a problem of handwriting recognition and a second treating a problem of open-set scene text recognition.
Handwriting recognition is the act of associating text with a written outline representing a letter, a word, sentence or a paragraph. This task remains challenging due to each person's different writing style and language-related difficulties to recognize. The current state-of-the art use deep learning models that need a lot of computing resources for the training. The project aims to design an efficient handwriting recognition model by extending the current state-of-the-art methods (Convolutional Recurrent Neural Network, Vertical Attention Network …) with new attention mechanisms. In addition to handwriting, we will be interested in handwriting that may contain erasures, subtexts and overwrites. We will start by focusing on the essays in Brazilian Portuguese and then recognize handwriting in other languages. The expected objectives at the end of the project are to improve performance in handwriting recognition, especially for text containing erasures.
Open-Set scene text recognition is an emerging task that models the constantly evolving character set in open-world character recognition applications. The current state-of-the art methods focus only on horizontal text. In this project, we propose to expand the existing open-set text recognition task to include vertical samples, resulting in a multi-orientation text recognition task. The expected objectives at the end of the project are to improve performance scene text recognition in open-set context with multiple orientations.