← Back to Publications

A Segmentation-Free Approach for Printed Devanagari Script Recognition

T. Karayil, A. Ul-Hasan, T. Breuel

IEEE International Conference on Document Analysis and Recognition (ICDAR)||DOI: 10.1109/ICDAR.2015.7333901
OCRDevanagariDocument AnalysisDeep Learning

Abstract

We propose a segmentation-free approach for recognizing printed Devanagari script, eliminating the need for character-level segmentation and enabling end-to-end recognition of Devanagari text.

Overview

Devanagari script recognition presents unique challenges due to the connected nature of characters and the complexity of the script. Traditional approaches rely on character-level segmentation, which is error-prone and limits recognition accuracy.

Segmentation-Free Approach

We propose a segmentation-free method that performs end-to-end recognition of printed Devanagari text without requiring explicit character segmentation. This approach simplifies the recognition pipeline and improves overall accuracy.