A novel method to auto configure convolution neural network model using soft computing technique to recognize telugu hand-written character for better accuracy

B.MEENA; K.VENKATA RAO; SURESH CHITTINENI

No Detail Found!

A novel method to auto configure convolution neural network model using soft computing technique to recognize telugu hand-written character for better accuracy

Author(s)

B.MEENA Department of Computer Science & Systems Engineering, Andhra University College of Engineering (A) , Andhra University and Raghu Engineering College, Visakhapatnam, Andhra Pradesh ,India

K.VENKATA RAO Department of Computer Science & Systems Engineering, Andhra University College of Engineering (A), Andhra University, Visakhapatnam , Andhra Pradesh ,India

SURESH CHITTINENI CSE Department , GITAM (Deemed to be University) , Department of Computer Science and Engineering, Visakhapatnam, Andhra Pradesh , India

Abstract

The paper presents the performance optimization using Genetic Algorithm with Convolutional Neural Networks (CNN) architecture to recognize hand written digits written in Telugu Language. CNN is tested with multiple configurations for various numbers of convolutional layers, filter size in each convolution layer, number of convolution filters in each layer, and pool size (for down sampling). the images in order to get optimal performance. Researchers have been using the trial-and-error approach of picking configurations for a CNN Model. This may not always guarantee the optimal performance and it needs, the user to monitor the performance trend on a regular basis with the improvement in prediction accuracy for changes in number of layers, filter size and number of filters. The Genetic Algorithm is used in this research to change the configuration of CNN to achieve the best results (accuracy of image recognition). Various architectures have been proposed by researchers for producing better results in Image recognition and classification areas. Our paper has proposed a method by changing CNN configurations with Genetic Algorithm and evaluated the overall test accuracy to 99% for hand written telugu characters.

Publication Details

Page(s) 5438-5447

DOI DOI not available

Published Journal: Journal of Theoretical and Applied Information Technology, Volume: 100, Issue: 18, Year: 2022

Request the full text from PASTIC

Keywords

Genetic Algorithm, convolution neural network, Soft Computing, Telugu Character Recognition

Citations

Downloads

125

Views