IMAGE CAPTION GENERATION SYSTEM FOR LOW-RESOURCE TELUGU LANGUAGE

Authors:

Ravi Gedam, Pushpa Birha, Pragati Agrawal, Mr.Saurabh Vyas, Abhishek Kundu

Keywords:

Image Caption Generation, Low- resource Language, Convolutional Neural Network, Flickr8K, Attention.

Abstract:

This paper introduces an innovative ap- proach to generating image captions for the low- resource Telugu language, an Indian language that is often overlooked. We address the scarcity of resources for the Telugu dataset by adapting the Flickr8K dataset through Google Translator, resulting in a custom Tel- ugu dataset. Despite Telugu’s cultural and regional significance, its limited resources pose challenges in advancing natural language processing, specifically in the domain of image caption generation. To accomplish the goal of caption generation, we employ the encoder- decoder framework by combining convolutional neural networks and recurrent neural networks. The ex- perimentation involves testing on the designed TICD dataset and demonstrates the notable performance in terms of BLUE and ROUGE scores.

PDF