Indian Art Form Recognition Using Convolutional Neural Networks

Abstract

Indian Culture is one of the richest cultures of the world. Art forms are a key component of the Indian culture that reveal a great deal about the various traditions, customs and practices of ancient as well as modern India. The paper proposes a framework classify Indian art forms into 8 different categories viz. Kalamkari, Kangra, Madhubani, Mural, Pattachitra, Portrait, Tanjore and Warli depending upon the style of art form. The proposed framework relies on the fusion of several state of the art deep convolutional neural networks for feature extraction from the dataset. Further, the experiments were carried out on a newly introduced dataset of over two thousand digital images of Indian paintings. The data-set is publically available for further experimentations. The model is able to achieve an accuracy of 86.56% outperforming other models.