Face Invariant Classification and Detection of Mythology Characters Using Custom Dataset (ClaDeMuC-CD)

Abstract

Innovation and learning are bound together. The digital space has evolved greatly with the innovations and new state-of-art-technologies where all domains adopt this dynamic change to explore and deliver to the people in many ways. A great interest thrived in us to explore the digital evolution in Indian mythologies. The basic notion is to connect the people with the mythological stories and to make it more innovative and creative to create a better and smarter world with manifold technologies available now. To give life to the stories, a few characters exist and people remember them forever a decade. Nowadays, it appears to be a challenging task for the youth to be aware of various mythological stories and the characters involved in those stories. The ignorance level of this kind of story seems to be more and it might not fascinate in the near future. Hence, to educate and create awareness about the mythological characters, the need for a learning model is required. The objective of this work is to deploy various models to do classification and detect Mahabharata characters like Arjun, Bheem, Karan, Krishna, etc. Moreover, experimental analysis has been done on state-of-the-art models like traditional Resnet50 and detection models like yolov5. Performance metrics such as training accuracies, validation accuracies, and then test accuracies are examined.