An Efficient Telugu Word Image Retrieval System using Deep Cluster

Abstract

Optical character recognition (OCR) system has successfully implemented in many applications of text classifications and text retrieval systems. But, due to the complexity of huge number of alphabets (symbols), grammatical punctuations,ithas failed to classify and retrieve many regional languages. Thus, to overcome these problems, word image retrieval is innovated. this article presents the Telugu word image retrieval(TWIR) system using advanced artificial intelligence based multi-layer deep learning convolution neural network (DL-CNN) is developed for extraction of accurate Telugu word features. Then, AlexNet based deep learning model is used for further clustering of features and the procedure named as DeepCluster. Thus, by using DeepCluster method logy various grammar rules of Telugu scripts will be perfectly analyzed. Hence, the feature database is trained with the Telugu word features along with its grammar rules. Thus, the system will effectively use for retrieval Telugu words and useful in all real time applications such as English to Telugu conversions and Telugu web browsing, Telugu speech analysis. The simulation results show that the proposed deep learning model gives the outstanding results compared to the state of art approaches for both Average Precision (mAP) and mean Average Recall (mAR) metrics.