Gajbhiye, Rajratan Laxminarayan (2024) Extraction of Devanagari handwritten characters using Deep Learning-based Models. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (1MB) | Preview |
Preview |
PDF (Configuration manual)
Download (1MB) | Preview |
Abstract
In several automation domains, such as Educational Technology, Digitization of Documents, The Analysis of Forensic Evidence, etc, Handwritten Character Recognition (HCR) is becoming more and more crucial. The approach of HCR involves the computer identifying and detecting each character in a text image and processing the data to create a machine-understandable format. The subject of recognition of patterns is a basic yet difficult job. In this research, we utilized the Devanagari Character Dataset. It is an open-source image dataset that contains 92,000 images of 46 different classes. This research investigates the efficiency and accuracy of three distinct models in training the recognition system: The proposed custom Convolutional Neural Network (CNN), Inceptionv3, and the Xception model. The proposed CNN approach is the most successful of these, obtaining an astounding accuracy of 99.11%. The results show that, out of all the models taken into consideration in this study, the Proposed custom CNN is the most accurate and computationally efficient model.
Actions (login required)
![]() |
View Item |