Abstract
Indian classical dances represent a significant aspect of the nation's cultural heritage, with each dance form characterized by its distinct style, posture, and movement. The automatic classification and recognition of these dances present considerable challenges due to differences in their postures, movements, and costumes. Recent advancements in deep learning techniques have yielded hopeful results in image classification tasks, including the classification of dance forms. In this study, a deep learning methodology using the MnasNet architecture is introduced for the classification of Indian classical dances. To enhance performance accuracy, the advanced polar fox optimization (APFO) algorithm is applied to optimize the hyperparameters of the MnasNet architecture. The experimental results are applied to Bharatnatyam Dance Poses dataset and its results have been compared with deferent state of the art similar models. Final results illustrate the efficacy of the proposed method in achieving elevated classification accuracy.