Crown-Centric Disguise Classification In Yakshagana Imagery Using Deep Learning Techniques

Abstract

Yakshagana, a theatrical art form in Karnataka, has variants like Thenku-thittu, BadaguThittu and Badaabadagu Thittu. Research explores its history, contemporary impact, and makeup trends. The Tenkutittu Yakshagana features different kinds of crowns. The type of crown the performer wears dictates the choice of disguise. For the purpose of disguise identification in this study, we have primarily examined classes such as kesaritatti, kolu kirita, pakdi, and kuttari. Therefore, this paper explores the use of deep learning methods for masking categorization in Yakshagana images, such as YOLOv5 and Three Tier CNN. A Cyclic Gate Recurrent Neural Network has been employed to classify Shaiva and Vaishnava characters. Following the categorization of the characters, the model is determining the disguise. Three-tier CNN classifies disguises with an accuracy of 85%. After conducting tests and assessments, it was concluded that YOLOv5, which has a 96% accuracy rate in identifying several items in an image, is the best appropriate algorithm for disguise categorization. Crown-Centric Disguise identification is a real-time tool that assists novices in determining which kind of crown and disguise a certain Yakshagana figure should wear.