Handwritten Changma Numerals Recognition Using Capsule Networks

Handwritten digit recognition is assumed to be a huge part in numerous authentication applications in the state-of-art technologies. As the manually written digits are not always found in similar size, thickness, style and orientation, recognition of handwritten digits is hardly possible in many cas...

Full description

Saved in:
Bibliographic Details
Main Authors: Hasan, Md Zahid, Hasan, K. M. Zubair, Hossain, Shakhawat, Mamun, Abdullah Al, Assaduzzaman, Md
Format: Conference Proceeding
Language:English
Subjects:
CapsulNet
Changma Handwritten Numerals Recognition
CNN
CVDR Dataset
Multi-Digits Recognition
Neural Network
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Handwritten digit recognition is assumed to be a huge part in numerous authentication applications in the state-of-art technologies. As the manually written digits are not always found in similar size, thickness, style and orientation, recognition of handwritten digits is hardly possible in many cases. To address this recognition tasks using handwritten digits, a great deal of work has been conducted with different image processing technologies on a variety of non-Indic's languages. However, this paper represents a handwritten digit recognition system of recognizing Changma Vaj Digit Recognition (CVDR). The digit recognition approach used in this paper is capsule network, an improved version of Convolutional Neural Network (CNN). The proposed system has been trained with more than 3440 sample image and been tested with more than 860 images. The proposed paper also presents a graphical comparison between the recognition accuracy performed by CNN and CapsNet, which in other terms demonstrates the supremacy of CapsNet's accuracy by 5.7%.
ISSN:2378-2692
DOI:10.1109/ICAEE48663.2019.8975468