Bharatanatyam hasta mudra categorization using deep learning approaches

The present study explores the precise identification of Bharatanatyam hasta mudras (hand gestures) by applying advanced convolutional neural networks and deep learning methodologies. Specifically, the article employs the Convolutional Sequential Neural Network, VGG-19, and, prominently, ResNet50V2 architectures to analyze a meticulously labeled dataset comprising Bharatanatyam hasta mudra images. With accuracy, recall, precision, and F1-score, key performance metrics are systematically employed for comprehensive performance evaluation, revealing an impressive accuracy rate of approximately 96.44%. Notably, the projected system is a valuable resource for Bharatanatyam practitioners, offering a sophisticated tool for discovering and mastering diverse hasta mudras. Its potential impact extends to dance schools and academies, where the system can be instrumental in teaching and integrating various hasta mudras into choreographic works. Beyond dance education, the article's contribution resonates in the broader context of preserving and promoting Indian classical dance forms.