Mythological Story Teller: Extending Scope Of Deep Learning To Indian Mythology

Abstract

Mythology holds immense importance as it serves as a repository of ancient wisdom, moral lessons, and cultural values. At the same time, Mythological Scene Recognition presents a unique challenge in Computer Vision as it demands precise identification and interpretation of characters and their relationship. Also, the lack of a dataset for detecting and identifying Mythological characters poses a significant obstacle. Subsequently, this paper presents a novel approach to Mythological Scene Recognition utilizing an integrated framework involving image processing, image recognition, web-based scrapping and a part of text analysis for summarization. The Mythological Scene Teller algorithm is proposed, which would fetch a picture of the past, identify the characters present in the scene, and give a summarized text representing the scene. A comparative study with different character recognition and classification models is also presented, treating the characters as objects. The paper incorporates a robust and efficient Web-scraping model to web scrape available information about the characters and establish relationships between the characters along with the integration of T5 Base for text summarization and information retrieval culminating an effective and enhanced approach.